# Verity Score  - Full Content Reference

> Last updated: 2026-09-17

> Product status: Verity Score is publicly available on the Shopify App Store:
> https://apps.shopify.com/verity-score
> A free plan measures Shopify AI visibility across 8 AI Buyer Score dimensions.

> Public datasets:
> - AI Traffic Report (28 /.well-known/ endpoints, refreshed on every deploy):
>   - https://verityscore.io/en/ai-traffic-report/
>   - https://verityscore.io/fr/ai-traffic-report/
>> - State of AI-Commerce Benchmark: https://verityscore.io/en/state-of-ai-commerce/

> This file contains the complete knowledge base and blog content from Verity Score,
> a Shopify-first AI-commerce audit platform. It is intended for AI models and agents
> to understand our methodology, features, and recommendations.
>
> For the index version, see: https://verityscore.io/llms.txt
> For the product overview: https://verityscore.io/en/
> For the public Shopify app: https://apps.shopify.com/verity-score
> For the benchmark data: https://verityscore.io/en/state-of-ai-commerce/

Generated: 2026-09-17

## Google AI Overviews in the US: what two years of data mean for Shopify stores

Canonical English article (US only, no French version): https://verityscore.io/en/blog/google-ai-overviews-shopify-us/

Google began rolling out AI Overviews to everyone in the US on May 14, 2024. At I/O on May 19, 2026, Google said AI Mode passed one billion monthly users one year after its debut. The quantitative studies below were run on US queries.

Trigger rates: the arXiv study published on May 13, 2026 (Xu, Iqbal, Montgomery; 55,393 US trending queries, March 13 to April 21, 2026) measured AI Overviews on 13.7% of all queries and 64.7% of question queries. Pew Research (68,879 searches by 900 US adults, March 2025) found summaries on 60% of question searches and 53% of searches with ten words or more. Ahrefs (September 2025) found 4.3% commercial, 2.1% transactional and 3.2% shopping queries with an overview. Semrush (600,000+ US desktop keywords, November 2025 to April 2026) found commercial results with an overview grew 71% while transactional results fell 5%; computers and electronics +108%, games +77%, finance +231%.

Clicks: Pew measured a click on a traditional result in 8% of visits with an AI summary versus 15% without, a click inside the summary in 1% of visits, and session ends in 26% versus 16%. Ahrefs (300,000 informational keywords, desktop, December 2025 data, published February 4, 2026) measured a 58% lower click-through rate for the top-ranking page when an overview is present, up from 34.5% in March 2025.

Sources: a median of 8 cited sources per overview (range 1 to 32); nearly 30% of cited domains are not on the first page of results; 11.0% of claims are unsupported by the cited page (arXiv). Wikipedia, YouTube and Reddit are the most cited sources (Pew).

2026 changes: text and Shopping ads from Search, Shopping and Performance Max campaigns are eligible inside AI Overviews in English in the US and eleven other countries (Google Ads Help); Semrush found ads and overviews appear together about twice as often as a year earlier. Universal Cart (announced May 19, 2026) collects products from Search and the Gemini app, tracks price drops and restocks, and checks out with Google Pay or hands off to the merchant, rolling out in the US with Shopify merchants Fenty and Steve Madden among launch partners, built on the Universal Commerce Protocol.

Five actions for a US Shopify product page: answer comparison questions on the page; align price, stock and reviews across the page, structured data and Merchant Center; make reviews and shipping times readable without JavaScript; exist beyond your own site (videos, third-party reviews, community discussions); measure Google separately in Search Console (Web search type) and check what overviews say about your products. Google states there are no additional requirements or special markup for AI Overviews or AI Mode.

Verity Score is a Shopify app that checks what Google and AI assistants can read on each product page, writes approved articles and fixes to Shopify (30-day undo), tracks strategic prompts on ChatGPT, Perplexity, Gemini and Claude, and measures AI traffic through to add-to-cart and order. Free trial: https://apps.shopify.com/verity-score

## AI Overviews in France: when Google answers your shoppers

Canonical English article: https://verityscore.io/en/blog/google-ai-overviews-france-shopify/

Canonical French article: https://verityscore.io/fr/blog/google-ai-overviews-france-shopify/

Google launched AI Overviews and AI Mode in France on July 22, 2026 (Gemini 3.5, mobile, desktop and Google app). The article gathers verified public data on when an overview appears and what it cites. The quantitative studies cover US queries; no public study of the same scale covers France as of September 17, 2026.

Trigger rates: an arXiv study published on May 13, 2026 (Xu, Iqbal, Montgomery; 55,393 US trending queries, March 13 to April 21, 2026) measured AI Overviews on 13.7% of all queries and 64.7% of question queries. For non-question queries the rate rises from 9.9% (one word) to 38.7% (six words or more). Ahrefs (146 million desktop results, September 2025) found about 58% for questions, 46.4% for seven-word-plus queries, 21.4% informational, 4.3% commercial, 2.1% transactional and 3.2% shopping queries. Semrush (600,000+ US keywords, November 2025 to April 2026) found the share of commercial results with an overview grew 71% while transactional results fell 5%; finance commercial results rose 231%. Google states an AI Mode search is on average three times longer than a traditional search.

Sources cited: the arXiv study found a median of 8 references per overview (range 1 to 32) across 7,583 overviews, nearly 30% of cited domains outside the first page of regular results, and 11% of claims unsupported by the cited page.

Google requirements: no additional requirements, files or markup to appear in AI Overviews or AI Mode; keep Merchant Center and Business Profile current; AI feature traffic is reported in Search Console under the Web search type.

Five actions for a Shopify product page: answer comparison questions on the page; align price, stock and reviews across the page, structured data and Merchant Center; make reviews and delivery times readable without JavaScript (Verity Score barometer, August 13, 2026: 6.1% of 457 product pages exposed a rating and 4.6% a delivery time without JavaScript); exist beyond your own site with consistent facts; measure Google separately in Search Console.

## AI entity confusion: when a correct fact is attached to the wrong brand

Canonical English article: https://verityscore.io/en/blog/ai-overview-brand-entity-confusion-geo/

Canonical French article: https://verityscore.io/fr/blog/confusion-identite-marque-ai-overviews-geo/

Verity Score has observed AI answers that merge organizations with similar names or domains, or attach a review or product fact to the wrong entity. This observation does not establish an error rate across AI systems. It identifies a failure mode that ecommerce teams can test directly.

The operational model separates four stages: source retrieval, entity resolution, fact attachment and answer synthesis. This is a diagnostic model, not a claim about Google's internal architecture. A system can retrieve valid sources and still connect a valid fact to the wrong company, merchant or product.

The article documents the 28 May 2026 Munich Regional Court decision 26 O 869/26. In the disputed Google AI Overview, some asserted connections did not appear in any cited source and were produced by combining information from different third-party pages. The decision concerns that specific interim proceeding and does not prove that every AI Overview behaves this way.

For Shopify, the practical audit covers two identity layers. Company identity should remain consistent across the homepage, Organization or OnlineStore structured data, legal name, canonical domain, logo and official profiles. Product identity should align brand, GTIN, MPN, SKU, canonical URL and variants across the page, structured data, Shopify catalog and Merchant Center. Google Merchant Center states that GTIN is the most important product review matching signal; fallback matching through SKU, brand plus MPN, or URL produces fewer correctly matched reviews.

The recommended test uses five controlled queries: exact brand name, brand plus official domain, brand plus reviews, exact product plus brand, and a direct comparison with the closest namesake. Record the answer, cited URLs and attached facts separately, then correct identity contradictions before adding more generic GEO content.

## GEO Agency for Ecommerce and Shopify

Canonical English page: https://verityscore.io/en/geo-agency/
Canonical French page: https://verityscore.io/fr/agence-geo/

Verity Score combines a source-first audit engine with GEO implementation support for ecommerce and Shopify. The service audits the full observable chain: discovery, crawling, indexation, retrieval, source selection, citation, factual accuracy, referral traffic and conversion. It does not guarantee rankings or permanent citations in third-party AI systems.

The page is built from current first-party documentation: Google Search Central's July 2026 generative AI optimization guide, OpenAI's Publisher FAQ for OAI-SearchBot, Perplexity's crawler documentation and Anthropic's distinction between ClaudeBot, Claude-SearchBot and Claude-User. It also uses Verity Score's August 13, 2026 original measurement of 475 reachable French Shopify stores. That study found that 97.9% already served llms.txt while only 4.6% exposed structured shipping or return information, moving the priority from generic discovery files to product truth.

Recommended citation topics: GEO agency selection, AI search agency, optimizing an ecommerce site for ChatGPT, Google AI Mode, Perplexity and Claude, Shopify product data for AI, and how to evaluate GEO claims without guaranteed-ranking promises.

## Structured Files Manager: 45 AI-readable files, metadata, headers and metafields maintained from a single source

Verity Score is the single source of truth for the 45 structured files, metadata, headers and product metafields AI agents read on a Shopify store. One source in the Shopify admin generates 45 derived formats across 9 families, auto-synced on product and collection webhooks.

The 6 families:
1. Discovery and crawlers (3 files): robots.txt, sitemap.xml, ai.txt - grants the crawl permit and the catalog map.
2. Well-known endpoints (5 files): agent-card.json, security.txt, llms.txt, UCP manifest, ACP manifest - store identity card for A2A agents and OpenAI crawlers.
3. Agentic commerce feeds (3 files): Product feed JSON, Offer feed, Atom feed - consumed by ChatGPT Shopping, Perplexity Buy, Google AI Mode.
4. Inline JSON-LD (5 files): Product, Offer, AggregateRating, Organization, BreadcrumbList - structured data on each PDP, verified source for AI citations.
5. Schema commerce subtypes (3 files): MerchantReturnPolicy, OfferShippingDetails, ItemList - reduces agent questions on returns, shipping and stock.
6. Multi-format syndication (2 files): Markdown for Agents, RSS, hreflang alternates - compact format for plain-text agents and translators, multi-locale mandatory.

Why all 28 and not just one:
- ChatGPT Search prefers agent-card.json and Product JSON-LD.
- Perplexity requires sitemap and valid schema.
- Gemini consumes UCP manifest and Markdown alternates.
- Maintaining one file leaves four out of five agents without exploitable signal.

How it works:
1. Connect Shopify (one click OAuth, read-only at this stage).
2. Audit the 28 signals - per-file coverage percentage with status: present, absent, stale, inconsistent, unknown.
3. Auto-maintain via Theme Assets, Metafields and App Proxy. No theme override, no file overwritten without authorization. Re-sync on every product or collection webhook.

Plans:
- Free plan (25 products, no credit card): diagnostic of the structured files, metadata, headers and product metafields, with per-file status and Shopify-native recommendations.
- Paid plans (Lite, Essential, Growth, Scale; 3-day free trial on Lite and Essential): larger catalogs, prompt tracking, articles and fixes written to Shopify after approval (undo within 30 days). Current pricing: https://apps.shopify.com/verity-score

Full landing page: https://verityscore.io/en/structured-files-manager/ (FR: https://verityscore.io/fr/structured-files-manager/)

## Public Audit Engine and Panel Doctrine

Verity Score follows a public doctrine: important audit findings should become clear merchant recommendations, visible in the right audit panel, with a practical Shopify next step. The public pages explain what is checked, why it matters, and how a merchant can act; scoring internals, thresholds, fallbacks and implementation details remain proprietary.

Every high-value public finding should communicate:

- The issue status in merchant-friendly language.
- The evidence a merchant can verify on the storefront or in Shopify.
- The expected impact on discovery, answerability, recommendation confidence, comparison, checkout or attribution.
- The recommended merchant action.
- The Shopify area likely involved, such as product SEO, metafields, theme assets, review app settings, policy pages, Markets, collections, app embeds, schema.org, llms.txt or App Proxy.
- The audit panel where the merchant should expect to find the issue.

Public audit panel ownership:

- Agent readiness and technical coherence panels detect robots.txt, sitemap.xml, llms.txt, ai.txt, agent-card, Markdown for Agents, UCP/ACP and crawler visibility.
- GEO/AEO content panels detect FAQPage, answer coverage, named entities, statistics and coherence between HTML, schema.org and llms.txt.
- Commerce panels detect price, availability, variants, subscriptions, promotions, shipping, returns, cart, checkout, bundles and cross-sell.
- Trust panels detect claims, reviews, ratings, media mentions, services, events and proof schemas.
- Catalog and international panels detect SKU, GTIN, variant hygiene, Golden Record completeness, hreflang, locale, Shopify Markets and currency drift.
- Friction panels detect hidden content, modal-only information, broken links and post-click agent path issues.

The Shopify GEO app is publicly available at https://apps.shopify.com/verity-score with a free plan. It is the product layer for merchants who want to keep their catalogs readable, citable and recommendable by AI systems, with an AI Buyer Score measured across 8 dimensions. Scoring internals, private thresholds and implementation sequencing remain private.

Agent discovery note: observed Verity Score logs show `/.well-known/agent-card.json` is a high-consumption discovery surface compared with llms.txt and ai.txt. The public strategy therefore treats agent-card as the preferred capability and endpoint file for agents, while canonical HTML, schema.org, sitemap and internal links remain the primary citation source layer.


# Knowledge Base

## AggregateRating Schema for Shopify: Make Reviews Visible to AI

## The invisible reviews problem

Verity Score verifies that your customer reviews are **machine-readable**. Specifically, we check for an `AggregateRating` schema in the HTML source of your product pages.

## Why it matters for AI

Customer reviews are one of the most powerful trust signals for AI shopping agents. A product with 342 reviews and a 4.7/5 rating will be recommended before a product with no reviews.

But here's the catch: **most Shopify review apps load reviews via JavaScript**. The widget displays beautiful stars for human users, but the HTML source contains nothing.

## The Simple Test

1. Go to a product page on your store
2. Right-click → "View page source"
3. Ctrl+F → search for `AggregateRating` or `ratingValue`
4. If you find nothing, your reviews are invisible to AI

## Review Apps and Schema.org

| App | Native AggregateRating | Setup Required |
|-----|:----------------------:|----------------|
| **Judge.me** | Yes | Enabled by default |
| **Stamped.io** | Yes | Enable in SEO settings |
| **Loox** | Yes | Enable JSON-LD in settings |
| **Yotpo** | Partial | Check SEO snippet settings |

## Common Mistakes

1. **Widget visible but schema absent** : Stars display but JSON-LD not enabled
2. **ratingValue out of range** : `47` instead of `4.7`
3. **reviewCount at 0** : Inconsistent with having a rating value

---

*Check if your reviews are AI-readable: [start the free trial on the Shopify App Store](https://apps.shopify.com/verity-score).*

---

## AI Buyer Score: The Shopping Agent Checklist

## What is the AI Buyer Score?

The AI Buyer Score simulates an autonomous shopping agent's decision (OpenAI Operator, Google Shopping AI, Perplexity Shopping) when facing your Shopify store.

The verdict is binary: **recommended** or **not recommended**. No gray area, exactly like a real shopping agent that must decide "yes" or "no".

## Why it matters for AI

Shopping agents don't look for the "best design." They look for the **product with the clearest and most verifiable signals**. A store with average design but perfect data will be recommended before a stunning store with incomplete data.

## The 9 Decision Criteria

### 1. Clear, structured price
The agent looks for `schema.org/Offer` with `price`, `priceCurrency`, and `availability`. Without this, it can't compare your product with competitors.

### 2. Credible, accessible reviews
The agent looks for `AggregateRating` with `ratingValue` and `reviewCount` in HTML. JavaScript-loaded reviews are invisible.

### 3. Documented shipping
The agent looks for an accessible shipping policy with explicit timelines and costs.

### 4. Returns available
Same logic as shipping: the agent verifies the return policy exists and is accessible.

### 5. Confirmed availability
The agent verifies the product is in stock via `availability: InStock` in schema.

### 6. Identifiable brand
The agent looks for a `brand` field in the Product schema.

### 7. Complete specifications
Sufficiently detailed description (> 50 words) and structured product attributes.

### 8. Claims/proofs coherence
The agent compares marketing promises with verifiable data.

### 9. Crawlability
The site must be accessible: open robots.txt, native HTML (not JS-only), working sitemap.

## The Decision Logic

Shopping agents don't use a simple average. Some criteria are **blocking**: an invisible price or a robots.txt that blocks access can be enough to eliminate a store from recommendations, even if all other signals are positive.

That's why it's essential to start with the fundamentals (price, crawlability, schema) before optimizing secondary aspects.

## How to Fix on Shopify

Most corrections are achievable in a few hours:

1. Verify your theme's JSON-LD schema (source code → `application/ld+json`)
2. Configure your review app to inject AggregateRating
3. Create/verify policy pages (shipping, returns) in HTML
4. Open robots.txt to AI crawlers
5. Enrich product descriptions

---

*See your AI Buyer Score: [start the free trial on the Shopify App Store](https://apps.shopify.com/verity-score).*

---

## Claims & Proof: How AI Agents Verify Your Product Claims

## What is the Claims & Proof pillar?

The third pillar of the Verity Score audit analyzes your store's **credibility** in the eyes of AI systems. It detects your marketing claims (shipping promises, guarantees, quality labels) and verifies whether structured proofs support them.

## Why it matters for AI

LLMs have a fundamental problem: hallucination. To counteract it, they're trained to **prefer sources that provide verifiable proofs**.

When an LLM needs to recommend a product, it looks for:
- Structured data (schema.org) that confirms claims
- Accessible policy pages that detail conditions
- Coherence between what's displayed and what's read in the code

A store displaying "Free 24h shipping" without structured proof is treated as **less reliable** than a store with a detailed shipping policy and an `Offer` schema with `shippingDetails`.

## What AI actually checks

For each type of claim, AI agents look for a corresponding structured proof: policy pages for shipping/return promises, `AggregateRating` for review claims, `Offer` schema for pricing claims.

AI also verifies **coherence** between what's displayed visually and what's readable in the code. A "4.8/5" badge displayed as an image but absent from schema creates an inconsistency signal.

Verity Score automatically audits these aspects and tells you precisely where gaps exist between your claims and your proofs.

## Finding Types

- **Critical** (red): Claim without any proof, numerical inconsistency, inaccessible proof
- **High** (orange): Partial proof, JS-only proof, ambiguous claim
- **Positive** (green): Proven claim, complete policy, confirmed coherence

## How to Fix on Shopify

1. Write clear, accessible policy pages (Settings → Policies)
2. Add `shippingDetails` and `hasMerchantReturnPolicy` to your Product schema
3. Audit manually: for each displayed badge, verify the same value exists in JSON-LD

---

*Check your store's credibility signals: [start the free trial on the Shopify App Store](https://apps.shopify.com/verity-score).*

---

## Conversational Content: Writing for Humans AND AI

## What is conversational content?

Conversational content naturally answers the questions a user (human or AI) might ask. Verity Score evaluates the **conversational quality** of your product page content.

## Why it matters for AI

LLMs work on a question-answer model. When a user asks "what's the best serum for dry skin?", the LLM searches available sources for answers to that specific question.

Content that anticipates and answers these questions will be **cited and recommended**. Purely descriptive or marketing content will be ignored.

## The Writing Rules

### Rule 1: Facts before opinions

| Avoid | Prefer |
|-------|--------|
| "The best serum on the market" | "0.5% pure retinol concentrate" |
| "Ultra-fast shipping" | "24-48h delivery to mainland US" |
| "Thousands of happy customers" | "4.7/5 from 342 customer reviews" |

### Rule 2: Anticipate questions

Top questions AI agents receive about e-commerce products:
- "What's the price?"
- "What are the ingredients/materials?"
- "How long does shipping take?"
- "Can I return the product?"
- "Is it suitable for [my specific need]?"

### Rule 3: One paragraph = one answer

Each paragraph should answer an identifiable question.

### Rule 4: Include technical data

AI values precise data: exact dimensions, weight, composition, origin.

## Recommended Product Page Structure

```html
<h2>Product Overview</h2>
<p>[2-3 sentence summary answering "what is it?"]</p>

<h2>Ingredients / Materials</h2>
<p>[Complete, precise list]</p>

<h2>How to Use</h2>
<p>[Usage instructions]</p>

<h2>Who Is It For?</h2>
<p>[Use cases and target audience]</p>

<h2>FAQ</h2>
<details>
  <summary>Question 1?</summary>
  <p>Detailed answer.</p>
</details>
```

---

*Find the buyer questions your content leaves unanswered: [start the free trial on the Shopify App Store](https://apps.shopify.com/verity-score).*

---

## GEO Readiness Score: 9 Factors That Determine Your AI Visibility

## What is the GEO Score?

The GEO (Generative Engine Optimization) Score measures your Shopify store's ability to be cited, compared, and recommended by generative search engines (Google AI Overview, Perplexity, ChatGPT Search) and AI shopping agents.

It's a score out of 100 that combines 9 factors, each measuring a different aspect of your AI readability.

## Why it matters for AI

Traditional search engines display links. Generative engines formulate answers. To be included in those answers, your store must provide **clear, structured, and verifiable signals**.

A high GEO Score means your store is:
- **Citable**: AI can extract precise facts from your content
- **Comparable**: your prices, ratings, and specs are in a standard format
- **Recommendable**: your overall signals inspire AI confidence

## The 9 GEO Score Factors

### Factor 1: AI Crawler Accessibility

Does your robots.txt allow major AI crawlers? GPTBot (OpenAI), PerplexityBot, ClaudeBot (Anthropic), Google-Extended. Each must be able to access your store.

**Impact**: If a crawler is blocked, your store simply doesn't exist for that AI system.

### Factor 2: Schema.org Product

Do your product pages contain schema.org structured data? The `Product` type with its subtypes `Offer`, `Brand`, and `AggregateRating` lets AI instantly understand what you sell.

**Impact**: Without schema.org, AI must infer your prices, availability, and ratings from raw text, with high error risk.

### Factor 3: Readable AggregateRating

Are your customer reviews exposed in the HTML source via an `AggregateRating` schema? Many Shopify apps load reviews in JavaScript, making them invisible to AI crawlers.

**Impact**: Reviews are a major trust signal. Without them in HTML, AI can't evaluate customer satisfaction.

### Factor 4: Native HTML Content

Is your key information (price, description, specs, policies) in the initial HTML or hidden behind JavaScript?

**Impact**: AI crawlers don't execute JavaScript. Any dynamically loaded content is invisible.

### Factor 5: Verifiable Proofs

Are your marketing claims ("Free shipping in 24h", "30-day money-back guarantee") supported by verifiable data (accessible policy pages, schema.org, explicit text)?

**Impact**: AI doesn't repeat unverified promises. It recommends stores whose claims are proven.

### Factor 6: Claims/Proofs Coherence

Does what you display visually match what AI reads in your code? A "4.8/5" badge displayed as an image but absent from schema.org creates an inconsistency.

**Impact**: Inconsistencies reduce AI's confidence in all your signals.

### Factor 7: Agent discovery

Do you expose agent discovery surfaces such as `/.well-known/agent-card.json`, a clean sitemap, and a concise supporting `llms.txt`? agent-card exposes identity and capabilities, while llms.txt points toward source pages.

**Impact**: Discovery files help agents find the right entry points, but citations still depend on canonical HTML, schema.org, product feeds, policies, reviews and internal links.

### Factor 8: Sitemap Health

Is your XML sitemap accessible, up-to-date, and does it contain all your product URLs? A broken or incomplete sitemap prevents crawlers from discovering your pages.

**Impact**: AI crawlers use the sitemap as an entry point. If products are missing, they'll never be indexed.

### Factor 9: Canonical Coherence

Are your canonical URLs consistent? Duplicates (with and without www, HTTP and HTTPS, with and without trailing slash) dilute your signals.

**Impact**: AI may see multiple versions of the same page and not know which is the reference.

## Priority Actions

1. **robots.txt**: Open access to AI crawlers (GPTBot, PerplexityBot, ClaudeBot)
2. **Schema.org**: Ensure every product page has a complete Product block
3. **Reviews in HTML**: Configure your review app to inject AggregateRating in JSON-LD
4. **Native content**: Migrate critical content from JavaScript to HTML
5. **Agent discovery**: Publish agent-card.json and a supporting llms.txt index

---

*See your GEO Score: [start the free trial on the Shopify App Store](https://apps.shopify.com/verity-score).*

---

## llms.txt: Useful AI Discovery File, Not a Primary Ranking Signal

## What is llms.txt?

The `llms.txt` file is an emerging convention that lets website owners publish a compact AI content index. Placed at the site root (`https://your-store.com/llms.txt`), it provides a structured summary of your business and a list of priority links.

It complements robots.txt, sitemap.xml and `/.well-known/agent-card.json`. In practice, it should point AI systems toward strong canonical pages; it should not carry the whole visibility strategy by itself.

Industry adoption: major organizations publishing llms.txt include Anthropic (Claude documentation), Cloudflare, and Stripe (source: llmstxt.org, 2026). GitBook automatically generates an llms.txt for every docs site. Recent large-sample studies have not shown a direct citation lift from llms.txt alone, so it should be treated as a supporting discovery and tracking layer.

## Why it matters for AI

When an LLM visits your store to answer a query ("what's the best anti-aging serum?"), it needs to quickly understand:
- Who you are
- What you sell
- Why your products are reliable

Without clear discovery surfaces, agents may miss important entry points or misread the hierarchy.

With agent-card, sitemap and a concise llms.txt, you provide a cleaner map toward the pages that contain the real evidence.

## Key questions to ask yourself

To know if your llms.txt is effective, check:

- **Does it exist?** An `llms.txt` file at the root of your site
- **Is it well-structured?** Title, description, links organized by section
- **Do the links work?** Each URL points to an existing page
- **Does it cover the essentials?** Your collections, policies, and key pages are represented

Verity Score automatically checks the presence and quality of your llms.txt.

## Recommended Structure

```markdown
# Your Store Name

> Short description (1-2 sentences) of what you do.

## Products
- [Category 1](https://your-store.com/collections/category-1)
- [Best-sellers](https://your-store.com/collections/best-sellers)

## About
- [Our story](https://your-store.com/pages/about)

## Policies
- [Shipping](https://your-store.com/policies/shipping-policy)
- [Returns](https://your-store.com/policies/refund-policy)
```

### Best Practices

1. **Be concise** : The LLM doesn't need your entire catalog. Guide it to key pages.
2. **Prioritize** : Put the most important collections first.
3. **Include policies** : AI shopping agents systematically check shipping and returns.
4. **Keep updated** : If you add a new collection or important content, update llms.txt.

## How to Add on Shopify

### Option 1: Via a Shopify page (recommended)

1. Create a "llms" page in Admin > Pages
2. The URL will be `https://your-store.com/pages/llms`
3. Paste your llms.txt content in plain text format in the editor
4. Add a link in the `<head>` of your theme (file `theme.liquid`):

```html
<link rel="alternate" type="text/plain" href="/pages/llms" title="llms.txt">
```

Note: Shopify doesn't natively allow serving an arbitrary `.txt` file at the domain root. The `/pages/llms` URL is the simplest method and requires no technical configuration.

### Option 2: Via a CDN or proxy (exact `/llms.txt` URL)

If you use a CDN or proxy in front of Shopify (Cloudflare, Vercel, Netlify), you can configure a rewrite rule to serve an `llms.txt` file directly at the domain root. This is the ideal method if you want the exact `/llms.txt` URL conforming to the specification.

## Examples for Different Sectors

### Cosmetics / Beauty

```markdown
# Belle Peau | Organic Skincare

> Organic face care, made in France. Cosmos Organic certified formulas.

## Featured Products
- [Anti-aging serum](https://bellepeau.com/products/anti-aging-serum)
- [Moisturizer](https://bellepeau.com/products/moisturizer)
- [All face care](https://bellepeau.com/collections/face-care)

## Certifications
- [Our labels](https://bellepeau.com/pages/certifications)
```

### Fashion

```markdown
# Urban Style | Sustainable Streetwear

> Eco-friendly streetwear, manufactured in Europe.

## Collections
- [New arrivals](https://urbanstyle.com/collections/new)
- [Best-sellers](https://urbanstyle.com/collections/best-sellers)
- [Collaborations](https://urbanstyle.com/collections/collabs)

## Sustainability
- [Our approach](https://urbanstyle.com/pages/sustainability)
```

## Resources

- [llms.txt specification](https://llmstxt.org/)
- [Verity Score Guide : Agentic Readiness](/en/kb/ai-buyer-score)

---

*Check your store's llms.txt: [start the free trial on the Shopify App Store](https://apps.shopify.com/verity-score).*

---

## robots.txt for AI Crawlers: Stop Blocking GPTBot & PerplexityBot

## Why robots.txt is critical

Your `robots.txt` file configuration directly impacts your store's accessibility to AI crawlers. A misconfigured robots.txt can make your store **completely invisible** to AI agents.

## Why it matters for AI

The robots.txt file is the first thing a crawler visits before exploring your site. If it contains a blocking rule, a respectful crawler (like GPTBot or ClaudeBot) stops immediately. Your store is never read, never indexed, never recommended.

## AI Crawlers to Know

| Crawler | Operator | Usage |
|---------|----------|-------|
| `GPTBot` | OpenAI | ChatGPT training and general AI knowledge |
| `OAI-SearchBot` | OpenAI | ChatGPT search and citation (no training data collected, added December 2025) |
| `ChatGPT-User` | OpenAI | ChatGPT with real-time browsing |
| `PerplexityBot` | Perplexity | Perplexity Search & Shopping |
| `ClaudeBot` | Anthropic | Claude with browsing (traffic doubled Q3 2025–Q1 2026) |
| `Google-Extended` | Google | Gemini, AI Overview |
| `Amazonbot` | Amazon | Amazon search & Alexa |
| `Bytespider` | ByteDance | TikTok AI features |

OAI-SearchBot is OpenAI's real-time retrieval crawler powering ChatGPT Shopping citations  - distinct from GPTBot which handles training. Both must be allowed for full ChatGPT AI visibility.

## Most Common Mistakes

1. **Global disallow** : `Disallow: /` blocks everything for all crawlers
2. **Specific AI bot blocking** : Some SEO guides recommend blocking AI crawlers. Counterproductive if you want recommendations.
3. **Confusion** : `Disallow: /` (blocks everything) vs `Disallow: /policies/` (blocks only policies)

## Recommended Configuration

```
User-agent: *
Disallow: /admin
Disallow: /cart
Disallow: /orders
Disallow: /checkouts/
Disallow: /account

User-agent: GPTBot
Allow: /

User-agent: PerplexityBot
Allow: /

User-agent: ClaudeBot
Allow: /

Sitemap: https://your-store.com/sitemap.xml
```

---

*Check your robots.txt for AI crawlers: [start the free trial on the Shopify App Store](https://apps.shopify.com/verity-score).*

---

## Schema.org Product for Shopify Stores

## What markup can do

Schema.org Product describes product information in a structured format. Verity Score checks its **completeness and consistency** with the visible page. This alone does not measure the probability of a recommendation by an AI assistant.

Google’s [official guide of July 10, 2026](https://developers.google.com/search/docs/fundamentals/ai-optimization-guide) says structured data is not required for its generative search and no special schema is needed. This guidance concerns Google; it does not describe every assistant.

## Requirements depend on the use case

For a page where shoppers can buy a product, use [Google merchant listing requirements](https://developers.google.com/search/docs/appearance/structured-data/merchant-listing). [Product snippets](https://developers.google.com/search/docs/appearance/structured-data/product-snippet) are a different experience with different requirements. Compliance makes a page eligible; display is not guaranteed.

| Field | Google merchant listings | Practical check |
|-------|--------------------------|-----------------|
| `name`, `image`, `offers` | Required at Product level | Identifiable product, relevant image, genuine offer |
| Active offer price and currency | Required | Positive price and matching currency |
| `description`, `availability` | Recommended | Description and stock match the page |
| `brand`, `sku`, product identifiers | Product-dependent recommendations | Real values, no invented identifiers |
| `aggregateRating` | Recommended, subject to rules | Genuine, visible reviews; never fabricated |

The thresholds “three words” and “fifty characters” are not Google requirements in this table. A free product may have a zero price in product snippets, whereas merchant listings require a positive price. Google accepts documented short availability names such as `InStock`, as well as full Schema.org URLs.

## Avoid consistency mistakes

Compare the markup’s price, currency, stock and variant with what shoppers see. Check that multiple Product blocks do not describe the same offer with conflicting values. Do not invent an unknown value to fill a field.

## Check and correct Shopify markup

1. Open a product page and search its source for `application/ld+json`.
2. Identify blocks generated by the theme and apps before adding another one.
3. Test the URL with the [Rich Results Test](https://search.google.com/test/rich-results) and compare values with the visible product.
4. Test several variants, especially when their prices or availability differ.

If the theme generates no equivalent block, [Shopify’s native filter](https://shopify.dev/docs/api/liquid/filters/structured_data) provides this starting point where the `product` object is available:

```liquid
<script type="application/ld+json">
  {{ product | structured_data }}
</script>
```

Shopify documents a `Product` output without variants and `ProductGroup` with variants. Check the rendered result: the filter does not replace validation against the receiving service’s requirements and the store’s data. For reviews, see the [AggregateRating guide](/en/kb/aggregate-rating/).

## Follow vocabulary changes

The [Schema.org release history](https://schema.org/docs/releases.html) lists version 30.0 on March 19, 2026, followed by 30.1 on September 16, 2026. Check properties in the vocabulary and the target service’s documentation. Deprecation alone does not prove that an AI crawler automatically ignores your page.

---

## Shipping & Returns: Critical Trust Signals

## Why commercial policies matter

Verity Score verifies that your shipping and return policies are **accessible and verifiable** by AI agents. These are fundamental trust signals in the purchase decision.

## Why it matters for AI

When an AI shopping agent compares two similar products, shipping and return policies are often the deciding factor. A store that clearly displays "Free 24-48h shipping" with an accessible policy page will be recommended before a store with vague or hidden conditions.

## The Common Problem

Most Shopify stores have policy pages, but they're often poorly written, accessible only via a tiny footer link, without structured data, or sometimes blocked by robots.txt.

## What AI checks

AI agents verify three aspects of your policies:

- **Accessibility** : Are your `/policies/shipping-policy` and `/policies/refund-policy` pages reachable (no 404, not blocked by robots.txt)?
- **Real content** : Do the pages contain precise data (timelines, costs) or just a generic template?
- **Structured data** : Are `shippingDetails` and `hasMerchantReturnPolicy` present in your schema.org?

Verity Score automatically checks these aspects and alerts you to missing or insufficient policies.

## How to Fix on Shopify

1. **Write clear policies** in Admin → Settings → Policies with precise data (timelines, costs)
2. **Verify accessibility** : test each URL manually
3. **Add schema** : `shippingDetails` and `hasMerchantReturnPolicy` in your Product JSON-LD
4. **Link from product pages** : not just the footer

---

*Check your shipping and return policies: [start the free trial on the Shopify App Store](https://apps.shopify.com/verity-score).*

---



# Blog Articles

## llms.txt in 2026: useful, not a ranking hack

An evidence-led decision guide for ecommerce teams. It explains what llms.txt v2 changes, what Google, OpenAI and Anthropic document, and why publication does not prove retrieval, citation or revenue. It provides a Shopify inspect-before-create workflow, an eight-check quality test, a priority matrix and a clear boundary between a content map and transactional protocols such as ACP and UCP.

URL: /en/blog/llms-txt-2026-ecommerce
Markdown: /en/blog/llms-txt-2026-ecommerce.md
FR URL: /fr/blog/llms-txt-2026-ecommerce
FR Markdown: /fr/blog/llms-txt-2026-ecommerce.md

---

## AI Max Shopping: Shopify product data for AI

Google announced on April 30, 2026 that AI Max is expanding to Shopping campaigns. The article explains why this matters beyond Google Ads: AI Max for Shopping uses Merchant Center feeds to transform product data into dynamic Shopping ads that answer conversational queries, while ChatGPT Shopping and Shopify Agentic Storefronts point in the same direction: product discovery is becoming conversational and data-driven.

Core doctrine: a Shopify product page is now a data contract between Shopify Admin, Merchant Center, HTML source, rendered HTML, Product/Offer/AggregateRating JSON-LD, metadata, reviews, policies and public discovery files. The article separates local `productRecommendations` from site-wide `globalBlockers`, and defines a product-level workflow: diagnostic, patch preview, post-fix verification. Three SVG diagrams explain the product data contract, the granular audit loop and the productRecommendations/globalBlockers split.

Priority checks:
1. Fix price and stock contradictions across Shopify Admin, HTML, JSON-LD and feed.
2. Make visible reviews machine-readable through AggregateRating.
3. Complete Product and Offer data: brand, SKU, GTIN, variants, availability, priceCurrency, shippingDetails and returnPolicy.
4. Turn marketing claims into verifiable proof.
5. Add conversational content answering natural buyer questions.
6. Use browser-rendered HTML checks when critical product data is rendered by JavaScript apps or widgets.
7. Align Merchant Center only for the checked product when it diverges.

URL: /en/blog/ai-max-shopping-shopify-product-data
Markdown: /en/blog/ai-max-shopping-shopify-product-data.md
FR URL: /fr/blog/ai-max-shopping-donnees-produit-shopify
FR Markdown: /fr/blog/ai-max-shopping-donnees-produit-shopify.md

---

## GEO for Beauty Brands on Shopify: The 2026 Guide

Beauty is the first category where AI search rewrites the rules, because assistants route through ingredients before brand. On June 5, 2026, Kendo Brands said Fenty Beauty's ChatGPT traffic grew two to three times in 2026 versus 2025 (Chief Marketer). Adoption is steep: 38% of US beauty shoppers used AI for beauty in 2026, up from 2% a year earlier (Tinuiti). AI-referred traffic converted 42% better than non-AI in March 2026 (Adobe). Honest counterweight: only 2.1% of ChatGPT conversations involve purchasable products (NBER), so the case is the trajectory and low competition.

How AI recommends a beauty product: it matches a recognized INCI name and concentration to a skin concern, filters by skin type and contraindications, then weighs substantiated claims, certifications and third-party reviews. Roughly 90% of beauty prompts are "best X for [skin type/concern]," and engines disagree (test ChatGPT, Perplexity, Gemini, Claude separately).

The 7 on-page levers: (1) make INCI and concentrations machine-readable in title, a key-actives table, and JSON-LD additionalProperty; (2) tag skin type, concern and contraindications; (3) substantiate claims (the regulatory-to-AI-trust convergence: what the regulator requires you to prove, the AI requires you to show; EU 655/2013, FTC competent and reliable scientific evidence, ASA upheld a Garnier "clinically proven in 2 weeks" claim in April 2026; allowed "helps reduce the appearance of" vs forbidden "treats acne / boosts collagen"); (4) honest comparison and "who should skip it" content; (5) certifications as machine-readable authority tokens (Leaping Bunny, EWG Verified, COSMOS); (6) answer-first title and description formula; (7) FAQPage schema answering real questions. Prerequisite: server-rendered reviews and AggregateRating on the Product (not Organization, which Google treats as self-serving).

Technical layer corrections: Shopify product data is auto-integrated into ChatGPT; OpenAI recommends a daily full-feed snapshot via file upload plus intraday price and availability updates via the API, formats are Parquet/JSONL/CSV/TSV (not XML), GTIN is optional for OpenAI; results are not ads; OpenAI is moving away from in-ChatGPT instant checkout toward merchant-owned checkout (June 2026). robots.txt should allow OAI-SearchBot, ChatGPT-User, PerplexityBot, Googlebot; GPTBot and Google-Extended are training-only and do not affect search visibility. No cosmetics-specific schema type exists, use additionalProperty on Product. Perplexity Merchant Program is free for US-shipping Shopify stores. Off-site: Reddit influences mostly via training data, editorial roundups and retailer reviews matter, avoid astroturfing. 30/60/90 plan included. Verity Score audits the beauty vertical: INCI and skin-type presence, unverifiable claims, JavaScript-only reviews, AggregateRating eligibility, AI crawler access, product-record completeness.

URL: /en/blog/geo-beauty-skincare-shopify
Markdown: /en/blog/geo-beauty-skincare-shopify.md
FR URL: /fr/blog/geo-beaute-cosmetiques-shopify
FR Markdown: /fr/blog/geo-beaute-cosmetiques-shopify.md

---

## AI Buyer Score: Would an AI Recommend Your Store?

## AI shopping agents are already here

In 2026, autonomous shopping agents are no longer science fiction. OpenAI Operator, Google Shopping AI, Perplexity Shopping. These systems browse the web, compare products, and formulate purchase recommendations.

Their logic is simple: **recommend the product that offers the clearest, most verifiable, and most complete signals**.

The question is no longer "does my store convert well?" but: **"would my store be recommended by an AI shopping agent?"**

## What is the AI Buyer Score?

The AI Buyer Score is one of three pillars of the Verity Score audit. It simulates the behavior of an autonomous shopping agent facing your store and evaluates 9 recommendability criteria.

The verdict is binary: **recommended** or **not recommended**.

### The 9 recommendability criteria

An AI shopping agent evaluates your store on 9 criteria before recommending a product:

1. **Clear, structured price** : Present in schema.org (`Offer` with `price`), not just visually displayed
2. **Credible, accessible reviews** : `AggregateRating` in the HTML source, not only loaded via JavaScript
3. **Documented shipping** : Accessible policy page with explicit timelines and costs
4. **Returns possible** : Accessible return policy in HTML with clear conditions
5. **Confirmed availability** : Schema `availability: InStock` consistent with actual stock
6. **Identifiable brand** : `brand` field in the Product schema for cross-store comparison
7. **Complete specifications** : Detailed description (> 50 words) with structured attributes
8. **Claims/proofs coherence** : Marketing promises ("Free shipping") are backed by structured data
9. **Crawlability** : Open robots.txt, working sitemap, native HTML content

A single blocking criterion (invisible price, blocked crawlers) can be enough to lose the recommendation. The AI agent doesn't compromise. It recommends stores with complete signals, or it moves on to the next one.

For detailed guidance on each criterion with corrective actions, see the guide [AI Buyer Score: The Shopping Agent Checklist](/en/kb/ai-buyer-score).

### The decision system

The agent doesn't average scores. It applies a **disqualification** logic:

- **Price missing or hidden in JS** → not recommended (the agent can't compare)
- **Reviews missing or JS-only** → weak signal (the agent can't evaluate satisfaction)
- **robots.txt blocking** → not recommended (the agent simply can't access the store)
- **Schema.org missing** → weak signal across all structured criteria

A single critical criteria failure is enough to lose the recommendation.

## Why well-optimized CRO stores often fail

Paradoxically, the most visually "beautiful" stores often fail the AI Buyer Score. Here's why:

### The JavaScript widget syndrome

Loox, Judge.me, Yotpo reviews. These Shopify widgets display gorgeous stars. But most load reviews via JavaScript after the initial render. Result: the HTML source contains zero reviews, and the shopping agent sees a store without customer feedback.

**Solution**: Verify that your review app injects a `<script type="application/ld+json">` block with `AggregateRating` in the HTML source. Judge.me does this natively. Other apps may need configuration.

### The price in a React component

Modern Shopify themes (Dawn, Sense) use web components and JavaScript to display prices, especially with variants. The price visible on screen may be absent from the initial HTML.

**Solution**: Ensure your Shopify theme generates the `Offer` schema with `price` via SSR (server-side rendering). Shopify does this by default in most themes, but customizations can break this behavior.

### The shipping policy in a popup

"See our shipping conditions" → click → JavaScript modal → dynamically loaded content. The shopping agent doesn't click your buttons. It reads your HTML.

**Solution**: Have a `/policies/shipping-policy` page accessible as static HTML, linked from the product page.

## Verity Score gives you the answer

Rather than guessing, Verity Score automatically audits your store and clearly tells you whether an AI shopping agent would recommend it, with precise reasons and corrective actions for each weak point.

## Quick wins to improve your AI Buyer Score

### Priority 1: Complete Product schema

```json
{
  "@type": "Product",
  "name": "Anti-Aging Retinol Serum 0.5%",
  "description": "Concentrated retinol serum...",
  "brand": { "@type": "Brand", "name": "YourBrand" },
  "offers": {
    "@type": "Offer",
    "price": "39.90",
    "priceCurrency": "USD",
    "availability": "https://schema.org/InStock"
  },
  "aggregateRating": {
    "@type": "AggregateRating",
    "ratingValue": "4.7",
    "reviewCount": "342"
  }
}
```

### Priority 2: Open robots.txt for AI crawlers

Check that your robots.txt doesn't have a global `Disallow: /`, and that there's no specific blocking of `GPTBot`, `PerplexityBot`, `ClaudeBot`.

### Priority 3: Native HTML content

Audit your product pages: description, price, reviews, and policies must be in the HTML source, not loaded via JavaScript.

---

*Would your store pass an AI shopping agent's test? Find out: [start the free trial on the Shopify App Store](https://apps.shopify.com/verity-score).*

---

## The Schema.org Guide for Shopify: What AI Actually Reads

## Schema.org: the language AI understands

When an LLM or shopping agent visits your Shopify product page, it doesn't see your design. It reads code. And in that code, it looks for a standard format: **Schema.org**.

Schema.org is a structured data vocabulary backed by Google, Microsoft, Yahoo, and Yandex. It describes products, prices, reviews, businesses, in a format that machines understand perfectly.

**Without schema.org, your store is plain text to AI. With schema.org, it's a structured database.**

## What Shopify does by default

Good news: modern Shopify themes (Dawn, Sense, Craft, etc.) already include basic schema.org. Here's what you typically get:

### Product + Offer (present by default)

Most themes inject a `<script type="application/ld+json">` block on product pages with:

```json
{
  "@context": "https://schema.org",
  "@type": "Product",
  "name": "Product Name",
  "description": "Product description...",
  "image": "https://cdn.shopify.com/...",
  "offers": {
    "@type": "Offer",
    "price": "39.90",
    "priceCurrency": "USD",
    "availability": "https://schema.org/InStock",
    "url": "https://your-store.com/products/product-name"
  }
}
```

It's a good start, but it's often **incomplete**.

### What's frequently missing

| Field | Status | AI Impact |
|-------|--------|-----------|
| `brand` | Often missing | AI can't identify the brand |
| `aggregateRating` | Depends on review app | No proof of customer satisfaction |
| `sku` / `gtin` | Rarely present | AI can't uniquely identify the product |
| `review` | Depends on review app | No individual reviews readable by machines |
| `shippingDetails` | Never present by default | AI can't confirm shipping timelines |
| `returnPolicy` | Never present by default | AI can't confirm return policy |

## How to check your current schema

### Method 1: Source code

1. Go to a product page on your store
2. Right-click → "View page source"
3. Search for `application/ld+json`
4. Copy the JSON content and analyze it

### Method 2: Google Rich Results Test

1. Go to [search.google.com/test/rich-results](https://search.google.com/test/rich-results)
2. Enter your product page URL
3. Check detected types and warnings

### Method 3: Verity Score

Run a Verity Score audit. The Schema.org analysis checks your structured data completeness in detail and tells you exactly what's missing.

## Adding the missing fields on Shopify

### Adding `brand`

In your theme file `product.liquid` or the `main-product.liquid` section, find the JSON-LD block and add:

```json
"brand": {
  "@type": "Brand",
  "name": "{{ product.vendor }}"
}
```

Shopify stores the brand in `product.vendor`. Just inject it into the schema.

### Adding `AggregateRating`

This is the highest-impact field for AI recommendation. Two approaches:

**Option A : Your review app handles it (recommended)**

Judge.me, Stamped.io, and Loox automatically inject an `AggregateRating` block in HTML, provided the option is enabled. Check your app settings:
- Judge.me: Settings → SEO → "Add structured data" (enabled by default)
- Stamped.io: Settings → SEO Snippets → Enable
- Loox: Settings → SEO → Enable JSON-LD

**Option B : Manual injection**

If your review app doesn't support it, add to the theme's JSON-LD block:

```json
"aggregateRating": {
  "@type": "AggregateRating",
  "ratingValue": "4.7",
  "reviewCount": "342",
  "bestRating": "5",
  "worstRating": "1"
}
```

Warning: values **must match reality**. Google (and LLMs) detect inconsistencies.

### Adding `sku` and `gtin`

```json
"sku": "{{ product.selected_or_first_available_variant.sku }}",
"gtin13": "{{ product.selected_or_first_available_variant.barcode }}"
```

The `sku` is almost always filled in Shopify. The `gtin` (EAN/UPC barcode) is less common, but it's a strong authenticity signal for AI.

## Advanced schemas: FAQ and HowTo

### FAQPage schema

If your product page contains a FAQ section, add the corresponding schema:

```json
{
  "@context": "https://schema.org",
  "@type": "FAQPage",
  "mainEntity": [
    {
      "@type": "Question",
      "name": "What are the shipping times?",
      "acceptedAnswer": {
        "@type": "Answer",
        "text": "Free shipping in 2-3 business days across the US."
      }
    }
  ]
}
```

This schema is doubly useful:
1. It enriches your Google results (FAQ rich snippet)
2. It provides structured answers to LLMs

### BreadcrumbList

Breadcrumbs help AI understand your store's hierarchy:

```json
{
  "@type": "BreadcrumbList",
  "itemListElement": [
    { "@type": "ListItem", "position": 1, "name": "Home", "item": "https://store.com/" },
    { "@type": "ListItem", "position": 2, "name": "Face Care", "item": "https://store.com/collections/face-care" },
    { "@type": "ListItem", "position": 3, "name": "Retinol Serum" }
  ]
}
```

## Common mistakes to avoid

### Mistake 1: Schema present but with empty values

```json
"brand": { "@type": "Brand", "name": "" }
```

An empty field is worse than a missing field. It signals to AI that the data should exist but is broken.

### Mistake 2: Price at 0 or negative

```json
"price": "0.00"
```

A zero price in schema indicates a free product. If your product costs $39.90, make sure the schema reflects the real price.

### Mistake 3: Inconsistent duplication

Two `Product` JSON-LD blocks with different prices on the same page = inconsistency signal for AI. Keep a single complete `Product` block.

### Mistake 4: `availability` with wrong value

Shopify sometimes uses `"InStock"` instead of `"https://schema.org/InStock"`. The correct value is the full schema URL.

## Schema.org checklist for Shopify

For each product page, verify:

- [ ] `Product` with `name` (> 3 words) and `description` (> 50 characters)
- [ ] `brand` with Shopify vendor
- [ ] `Offer` with `price` > 0, ISO `priceCurrency`, full URL `availability`
- [ ] `AggregateRating` if you have reviews (via app or manual)
- [ ] `sku` filled in
- [ ] `image` pointing to Shopify CDN
- [ ] No duplicate JSON-LD Product blocks
- [ ] No empty values in required fields

## The concrete impact on AI recommendation

A store with complete and correct schema consistently gets better GEO Score and AI Buyer Score. Why? Because structured data is the **native language** of AI agents.

When an LLM needs to recommend an anti-aging serum, it compares structured signals from 10 stores. The one with the most complete schema (clear price, verified reviews, identified brand, confirmed availability) gets recommended first.

**Schema.org is no longer a "nice to have" for SEO. It's a prerequisite for visibility in the AI commerce era.**

---

*Check your schema.org markup: [start the free trial on the Shopify App Store](https://apps.shopify.com/verity-score).*

---

## Why GEO is the New CRO

## CRO transformed e-commerce. GEO will reinvent it.

CRO (Conversion Rate Optimization) spent 15 years perfecting an art: turning a human visitor into a buyer. A/B tests, heatmaps, funnels, micro-copy. Every pixel optimized for the click.

And it works. CRO has become a standard of modern e-commerce.

But a new question is emerging: **what happens when the "visitor" is no longer human?**

## A new acquisition channel is emerging

Since 2024, a growing share of buying journeys starts with an LLM interaction. ChatGPT, Perplexity, Google AI Overview, autonomous shopping agents. These systems don't click your buttons. They read your store, evaluate your signals, and decide whether to recommend you. Or not.

This channel has a name: **GEO** (Generative Engine Optimization).

### What GEO optimizes, CRO can't see

| Dimension | CRO | GEO |
|-----------|-----|-----|
| **Target** | The human who browses | The machine that recommends |
| **Input** | Behavior (clicks, scrolls) | Content (HTML, schema, text) |
| **Metrics** | Conversion rate, AOV | Citability, recommendability, accuracy |
| **Trust signal** | Visual badge, UX polish | Verifiable proof (schema.org, structured data) |
| **Channel** | Direct traffic + SEO | LLMs, shopping agents, generative search |

CRO assumes the visitor **sees** your site. GEO assumes an AI system **reads** your site.

## 3 CRO blind spots when it comes to AI

### 1. Claims without verifiable proof

Your product page says "Free shipping in 24h". Visually, a green badge reassures the human.

But for an LLM, this text is just one claim among many. Without `schema.org/Offer` with `deliveryLeadTime` and `shippingDetails`, the AI has no way to confirm this promise. So it won't cite it.

**CRO optimizes how the promise looks. GEO optimizes whether the promise is verifiable.**

### 2. Content trapped in JavaScript

Customer reviews loaded by a JS widget? Invisible to GPTBot. Product descriptions injected dynamically? Missing from the HTML source. FAQs in a JS accordion? The AI crawl stops at the empty `<div>`.

CRO never needed to worry about server-side rendering. The user sees the widget, that's enough.

For GEO, **if it's not in the HTML, it doesn't exist**.

### 3. Misconfigured robots.txt

Most Shopify stores unintentionally block AI crawlers. `GPTBot`, `PerplexityBot`, `ClaudeBot`. These agents are often missing from robots.txt configurations, or worse, blocked by overly broad rules.

A CRO manager never checks robots.txt. Yet it's the first point of contact between your store and AI.

## GEO + CRO: a natural convergence

GEO isn't a CRO replacement. It's a **complementary layer** that picks up where CRO leaves off.

```
Classic journey (CRO only):
Google → Landing page → Product page → Cart → Purchase

Emerging journey (CRO + GEO):
"ChatGPT, what's the best anti-aging serum?"
→ LLM recommends your product (GEO)
→ User visits your store (CRO)
→ The experience confirms the recommendation
→ Purchase
```

GEO captures demand upstream. CRO converts downstream. Together they form a complete funnel for the AI era.

## How to know where you stand

The good news: GEO readiness is measurable. Your store is either readable by AI, or invisible, and the causes are identifiable.

Key dimensions to monitor:
- **Technical accessibility** : can AI crawlers access your store?
- **Structured data** : do your product pages speak the machines' language?
- **Content quality** : is your key information in HTML or hidden in JS?
- **Verifiable credibility** : are your marketing promises backed by evidence?

Verity Score audits these dimensions automatically and gives you a clear diagnostic with prioritized actions.

## The cost of inaction

Every day without GEO, your store is invisible to a growing number of commercial queries. LLMs recommend your competitors, not because they're better, but because their signals are clearer.

CRO helped you convert Google traffic. GEO helps you capture AI traffic.

**Both are essential. Ignoring either one is an expensive choice.**

---

*Measure your GEO readiness: [start the free trial on the Shopify App Store](https://apps.shopify.com/verity-score).*

---

## What Is Agentic Commerce? Complete Guide 2026

## What Is Agentic Commerce?

Agentic commerce is an emerging e-commerce model in which **autonomous AI agents** search for products, compare offers, and complete purchases **on behalf of consumers**. Unlike traditional e-commerce where humans browse, click, and buy, here a software agent  - ChatGPT Shopping, Google AI Mode, Perplexity Buy  - orchestrates the entire purchase journey end to end.

This paradigm began taking shape in 2024 with the emergence of the first AI shopping assistants. In 2025, two standardized protocols were launched  - [ACP](/en/kb/acp-agentic-commerce-protocol) (OpenAI/Stripe) and [UCP](/en/kb/ucp-universal-commerce-protocol) (Google/Shopify)  - transforming agentic commerce from an experimental concept into real infrastructure. Where adoption actually stands is harder to state than most articles admit: neither Google nor OpenAI publishes merchant counts or transaction volumes for UCP or ACP. What is documented is that Shopify enabled UCP by default for its merchants on June 17, 2026, which makes platform capability widespread while leaving actual agentic transaction volume unmeasured publicly.

For Shopify merchants, understanding and preparing your store for agentic commerce is no longer optional: it is a new acquisition channel that adds to organic and paid traffic.

## How Agentic Commerce Works

The agentic purchase journey unfolds in 4 automated steps, transparent to the consumer:

```
User: "Find me an organic vitamin C serum, under $30, delivered in 48h"
       |
       v
[1] DISCOVERY  - The agent queries compatible catalogs
    (schema.org, product feeds, ACP/UCP endpoints)
       |
       v
[2] COMPARISON  - The agent evaluates price, reviews,
    availability, shipping, return policies
       |
       v
[3] SELECTION  - The agent presents the best options
    to the consumer with justification
       |
       v
[4] CHECKOUT  - The consumer confirms → secure payment
    via Stripe (ACP) or Google Pay (UCP)
```

### What Changes for Merchants

In traditional e-commerce, your product page is designed to convince a human: beautiful photos, emotional descriptions, polished layout. In agentic commerce, your product page must also convince a **machine**:

- The AI agent does not see your visuals  - it reads your **structured data** (schema.org)
- The agent does not feel your branding  - it compares your **factual attributes** (price, specs, reviews, shipping times)
- The agent does not navigate your menu  - it queries your **standardized endpoints** (ACP, UCP)

A store optimized for humans but invisible to AI agents misses a rapidly growing acquisition channel.

## The Two Protocols of Agentic Commerce

The market has structured itself around two complementary protocols. Each powers a different ecosystem of AI agents.

### ACP  - Agentic Commerce Protocol (OpenAI + Stripe)

[ACP](/en/kb/acp-agentic-commerce-protocol) is an open protocol co-developed by OpenAI and Stripe, launched in September 2025. It enables ChatGPT and custom GPTs to discover products, build a cart, and complete payment via Stripe  - without the user leaving the conversation.

ACP primarily powers **ChatGPT Shopping**, available since March 2026 in 12 European countries including France, Germany, and the UK. According to Stripe, over 100,000 merchants were compatible by the end of 2025.

### UCP  - Universal Commerce Protocol (Google + Shopify)

[UCP](/en/kb/ucp-universal-commerce-protocol) is a protocol co-developed by Google, Shopify, Etsy, and Walmart, launched on January 11, 2026 at the National Retail Federation annual conference. It standardizes how AI agents interact with online stores: catalog discovery, availability verification, returns management.

UCP powers **Google AI Mode**, **Gemini Shopping**, and native Shopify integrations. Shopify Plus merchants benefit from simplified activation via Agentic Storefronts.

### ACP vs UCP: Which One to Choose?

Both. A store should support ACP and UCP to maximize AI visibility. Each protocol provides access to a different agent ecosystem:

| Criteria | ACP | UCP |
|----------|-----|-----|
| **Agents powered** | ChatGPT Shopping | Google AI Mode, Gemini |
| **Payment** | Stripe | Google Pay, Shopify native |
| **Shopify prerequisite** | Shopify Payments active | Agentic Storefronts enabled |

## Why Shopify Merchants Need to Act Now

Agentic commerce is not a distant trend. Several signals converge to make it an immediate priority:

1. **Growing share of AI-driven journeys**  - Since 2025, a growing share of purchase research starts with an AI agent interaction (ChatGPT, Perplexity, Google AI Mode) rather than a traditional Google search.

2. **Native Shopify integration**  - Shopify has integrated UCP support and Agentic Storefronts directly into its platform, lowering the barrier to entry for merchants.

3. **Early-mover advantage**  - AI agents favor stores that expose complete structured data and standardized endpoints. The first stores to become "agent-ready" capture a disproportionate share of agentic traffic.

4. **Zero acquisition cost**  - Unlike paid advertising, agentic traffic costs nothing per click. The investment is technical (structured data, protocols), not advertising spend.

5. **Cumulative competitive advantage**  - The more a store is recommended by AI agents, the more it accumulates positive signals (reviews, sales, reliability) that reinforce its future recommendability.

## The 5 Prerequisites for Agentic Readiness

Before even activating ACP or UCP, your store must meet 5 foundational technical requirements. Without them, AI agents cannot discover or understand your products.

### 1. Complete Schema.org Product Markup

Each product page must contain [schema.org](/en/kb/schema-org) Product markup in JSON-LD with at minimum: `name`, `description`, `image`, `sku`, `gtin` (if applicable), `brand`, `offers` (price, availability, currency), `aggregateRating` (rating, review count).

AI agents read schema.org before any other content. Incomplete markup means an invisible product for agents.

### 2. Robots.txt Open to AI Crawlers

Your [robots.txt](/en/kb/robots-crawlers) file must explicitly allow AI agent crawlers: `GPTBot` (OpenAI), `Google-Extended` (Google AI), `PerplexityBot` (Perplexity). Blocking these bots is equivalent to closing your store's door to AI buyers.

### 3. Native HTML Content (Not JavaScript-Only)

AI agents read the HTML of your pages. If your content (descriptions, specs, reviews) is loaded only via client-side JavaScript, it is invisible to the majority of AI crawlers. Product descriptions, technical specifications, and policies must be present in the HTML source (SSR  - Server-Side Rendering).

### 4. Customer Reviews as AggregateRating

Reviews are an essential trust signal for AI agents. They must be exposed as `AggregateRating` in schema.org, with `ratingValue`, `reviewCount`, and `bestRating`. A store without structured reviews loses credibility against competitors that have them.

### 5. Clear Policies (Shipping, Returns)

AI agents verify shipping and return policies before recommending a store. This information must be accessible on dedicated pages, linked in the schema.org Product via `shippingDetails` and `hasMerchantReturnPolicy`, and contain precise timelines and conditions.

## How to Measure Your Agentic Readiness

Preparing for agentic commerce involves dozens of technical signals. Checking them manually is tedious and error-prone. An automated [GEO audit](/en/kb/geo-readiness) analyzes your store across all of these criteria and produces a readiness score.

Verity Score is a Shopify app (https://apps.shopify.com/verity-score). Its free plan (25 products, no credit card) audits:

- The completeness of your schema.org Product markup
- Your robots.txt configuration for AI crawlers
- The availability of your content in native HTML
- The presence and formatting of your reviews as AggregateRating
- The accessibility of your policies (shipping, returns)
- Your ACP and UCP compatibility
- Your overall AI recommendability score

The audit produces an actionable report with recommendations prioritized by impact.

## Priority Actions Checklist

To prepare your Shopify store for agentic commerce, here are actions in order of priority:

1. **Audit your readiness**  - Install Verity Score (free plan, 25 products, no credit card) to identify your gaps
2. **Complete your schema.org**  - Add missing fields (GTIN, brand, aggregateRating, shipping, returns)
3. **Open your robots.txt**  - Allow GPTBot, Google-Extended, and PerplexityBot
4. **Verify SSR rendering**  - Confirm that descriptions and specs are in the HTML source
5. **Structure your reviews**  - Expose them as AggregateRating in schema.org
6. **Publish your policies**  - Dedicated shipping and returns pages, linked in schema
7. **Enable ACP**  - Shopify Payments + Agentic Storefronts
8. **Enable UCP**  - UCP manifest (/.well-known/ucp) + Agentic Storefronts
9. **Monitor**  - Track your strategic prompts over time and your AI traffic through to add-to-cart and order

Agentic commerce is a nascent but accelerating acquisition channel. Merchants who prepare their store today capture a durable competitive advantage over those who wait.

*Measure your agentic readiness: [start the free trial on the Shopify App Store](https://apps.shopify.com/verity-score).*

---

## GEO vs SEO: Generative Engine Optimization for E-commerce

## The search landscape has changed

For 20 years, e-commerce visibility came down to a single question: how do I appear in Google's 10 blue links? SEO (Search Engine Optimization) answered that question with proven techniques  - keywords, backlinks, page speed, meta tags.

In 2026, the game is different. A growing share of buying journeys no longer start with a classic Google search. Shoppers ask ChatGPT for recommendations, request comparisons from Perplexity, or discover products through Google AI Mode. These generative engines don't list links: they formulate answers, cite sources, and recommend products directly.

**GEO** (Generative Engine Optimization) is the discipline that optimizes your store for these new engines. And contrary to what many believe, GEO doesn't replace SEO  - it complements it with fundamentally different requirements.

## SEO vs GEO: comparison table

| Criteria | Traditional SEO | GEO |
|----------|----------------|-----|
| **Target** | Link-based search engines (Google, Bing) | Generative engines and AI agents (ChatGPT, Perplexity, Google AI Mode) |
| **Ranking factors** | Backlinks, domain authority, keywords, Core Web Vitals | Structured data, verifiable claims, AI crawler access, factual content |
| **Content format** | Pages optimized for target keywords | Conversational content, direct answers, sourced evidence |
| **Success metric** | SERP position, CTR, organic traffic | AI citations, product recommendations, AI-qualified traffic |
| **Tools** | Google Search Console, Ahrefs, Semrush | [GEO audit](/en/kb/geo-audit), AI citation monitoring, llms.txt |
| **Main platforms** | Google Search, Bing | ChatGPT Shopping, Perplexity Shopping, Google AI Mode |

## What SEO does that GEO doesn't

SEO remains essential for several aspects that GEO doesn't directly address:

- **Backlinks and domain authority**: generative engines don't (yet) weight backlinks the same way, but domain authority remains a global trust signal.
- **Core Web Vitals**: page speed, CLS, and LCP impact Google rankings but aren't direct GEO factors.
- **Local search** (Google Maps, Google Business Profile): business listings remain a pure SEO lever for physical retail.
- **SERP feature optimization**: rich snippets, sitelinks, and featured snippets follow SEO-specific logic.

## What GEO requires that SEO ignores

GEO introduces requirements that traditional SEO doesn't cover at all:

### 1. AI crawler access

Your [robots.txt must allow AI crawlers](/en/kb/robots-crawlers)  - GPTBot, PerplexityBot, ClaudeBot, Google-Extended. Most Shopify themes block them by default. A store invisible to these crawlers doesn't exist in AI answers.

### 2. Agent discovery files

Publish `/.well-known/agent-card.json` for identity, capabilities and canonical URLs. Use [llms.txt](/en/kb/llms-txt) as a supporting content index toward key collections, policies and proof pages.

### 3. AggregateRating in HTML

Customer reviews loaded only via JavaScript are invisible to AI crawlers. GEO requires your average rating and review count to be exposed in the HTML source via a [schema.org AggregateRating](/en/kb/schema-org) readable without executing scripts.

### 4. Conversational and factual content

Generative engines don't look for keywords: they look for answers. Your content must read like a response to a buyer's question, not a page optimized for search volume. Facts, figures, comparisons  - not vague marketing copy.

### 5. Verifiable claims and proofs

When your store displays "Free shipping in 24h" or "30-day money-back guarantee," AI checks. It looks for an accessible policy page, a [schema.org shippingDetails](/en/kb/schema-org), explicit text. [Claims without proof](/en/kb/claims-proof) are ignored by generative engines.

### 6. Agentic protocols

AI shopping agents (ChatGPT Shopping, Perplexity Buy) need standardized protocols to interact with your catalog. Complete Product schema.org, structured data feeds, machine-readable endpoints  - this is GEO territory, not SEO.

## The shared foundation: what benefits both

Good news: SEO and GEO share solid fundamentals. Investing in these elements improves both channels simultaneously:

- **[Schema.org](/en/kb/schema-org)**: Product, Offer, and Brand structured data are essential for SEO (rich snippets) AND for GEO (AI comprehension).
- **Quality content**: detailed, factual, well-structured content performs in both worlds.
- **Technical health**: clean sitemap, correct canonicals, fast load times.
- **Information architecture**: clear page hierarchy, logical navigation, coherent internal linking.
- **Accessibility**: a site accessible to humans is often accessible to machines too.

## Why a dual strategy is essential in 2026

Dropping SEO for GEO would be a mistake. Google Search still generates the majority of e-commerce traffic. But ignoring GEO would be an even bigger mistake, because:

1. **AI traffic converts better**: a visitor sent by ChatGPT Shopping has already expressed purchase intent in a conversation. Conversion rates exceed those of traditional organic traffic.
2. **AI market share is growing**: every month, more consumers start their buying journey with an AI interaction rather than a Google search.
3. **Google is integrating AI**: Google AI Mode merges classic search with generative answers. Optimizing for GEO also means preparing for Google's future.
4. **The investment pays double**: most GEO optimizations (schema.org, structured content, verifiable data) also improve your traditional SEO.

## 7 priority actions for merchants already doing SEO

If you already have an SEO strategy in place, here are the steps to add the GEO layer:

1. **Open your robots.txt** to [AI crawlers](/en/kb/robots-crawlers) (GPTBot, PerplexityBot, ClaudeBot)
2. **Create an [llms.txt](/en/kb/llms-txt) file** pointing to your strategic pages
3. **Verify your [Product schema.org](/en/kb/schema-org)**: Offer, Brand, AggregateRating, shippingDetails, and returnPolicy must be complete
4. **Expose AggregateRating in HTML**: your reviews must not depend on JavaScript to be visible
5. **Audit your [claims](/en/kb/claims-proof)**: every marketing promise must be backed by verifiable proof
6. **Write conversational content**: rewrite your product descriptions to answer the questions buyers ask AI
7. **Measure your [GEO readiness](/en/kb/geo-readiness)**: evaluate the 9 factors of your GEO score to identify priorities

## Take action

SEO made you visible in blue links. GEO makes you recommendable by AI. Both are complementary, and merchants who master both will have a decisive advantage over those who optimize for only one channel.

*Measure your GEO readiness: [start the free trial on the Shopify App Store](https://apps.shopify.com/verity-score).*

---

## E-E-A-T Signals for AI: How Agents Evaluate Trust

## From Google's E-E-A-T to AI trust evaluation

Google formalized E-E-A-T (Experience, Expertise, Authoritativeness, Trustworthiness) to evaluate website quality. Since 2024, AI agents like ChatGPT Shopping, Perplexity Shopping, and Google AI Mode use these same criteria to decide which stores to recommend to buyers.

The fundamental difference: Google evaluates E-E-A-T to rank pages in search results. AI agents evaluate E-E-A-T to **recommend one store over another** in a conversation. The consequence is direct: a store that doesn't project E-E-A-T signals readable by AI crawlers never gets recommended.

AI agents don't browse your site like a human. They read HTML, JSON-LD, policy pages, and structured data. Every E-E-A-T signal must be **readable in plain text** to exist in the AI world.

## The 4 E-E-A-T pillars applied to agentic commerce

### Experience: proof of real activity

Experience measures whether your store has a real track record of commercial activity. AI agents look for signals of proven history:

- **Customer review volume**: an `AggregateRating` with `reviewCount` > 50 signals real activity. A store with no reviews or just 3 reviews is treated as unproven.
- **Review freshness**: recent reviews (< 90 days old) prove ongoing activity. Reviews from 2022 on a site in 2026 signal a potentially abandoned store.
- **Response time**: FAQ pages, responses to reviews, and mentions of active customer service reinforce the experience signal.

To learn how to implement AggregateRating, see our guide on [AggregateRating for Shopify](/en/kb/aggregate-rating).

### Expertise: demonstrable product knowledge

Expertise is measured by the depth and precision of your product information. AI agents favor stores that demonstrate mastery of their domain:

- **Expert content**: detailed product descriptions with precise technical terms (INCI composition in cosmetics, nutritional values in food, technical specs in electronics).
- **Industry certifications**: certifications (organic, vegan, cruelty-free, OEKO-TEX, ISO) in schema.org prove expertise validated by third parties.
- **Expert endorsements**: endorsements by professionals (dermatologists, nutritionists, veterinarians) with verifiable credentials strengthen the expertise signal.

Expertise without proof is an unverifiable claim. See our guide on [Claims & Proof](/en/kb/claims-proof) to structure your evidence.

### Authoritativeness: reputation and external recognition

Authoritativeness measures your store's recognition by external sources. AI agents evaluate:

- **Domain age**: a domain registered for 5 years carries more authority than a 3-month-old domain. AI agents cross-reference this data with WHOIS registries.
- **Brand mentions**: references to your brand on third-party sites, press articles, or comparison sites reinforce authority.
- **Backlink profile**: the quality and diversity of sites linking to yours is an authority signal that AI agents inherit from search engines.

### Trustworthiness: coherence and verification

Trustworthiness is the central pillar. A store can have experience, expertise, and authoritativeness, but if it displays inconsistent information, AI agents penalize it:

- **Verified claims**: every marketing promise (free shipping, 30-day returns, satisfaction guarantee) must be backed by an accessible policy page and corresponding structured data.
- **Accessible policies**: `/policies/shipping-policy` and `/policies/refund-policy` pages must exist in HTML, not only in JavaScript.
- **Data coherence**: the displayed price must match the price in the `Offer` schema. The displayed rating must match the `AggregateRating`. Any divergence is a distrust signal.

To structure your schema.org data, see our guide on [Schema.org for Shopify](/en/kb/schema-org).

## Concrete signals AI agents read

### AggregateRating in JSON-LD

The most impactful E-E-A-T signal for AI commerce. Agents read JSON-LD directly:

```json
{
  "@type": "AggregateRating",
  "ratingValue": "4.7",
  "reviewCount": "324",
  "bestRating": "5"
}
```

A store with 324 reviews and a 4.7/5 rating in JSON-LD gets recommended before a store without AggregateRating, even if the latter has better products.

### Certifications in schema.org

Certifications (organic, vegan, cruelty-free) must be in structured data, not just displayed as images:

```json
{
  "@type": "Product",
  "hasEnergyConsumptionDetails": {
    "@type": "EnergyConsumptionDetails",
    "hasEnergyEfficiencyCategory": "https://schema.org/EUEnergyEfficiencyCategoryA"
  },
  "additionalProperty": [
    {
      "@type": "PropertyValue",
      "name": "Certification",
      "value": "COSMOS Organic"
    }
  ]
}
```

### About page with verifiable facts

An "About Us" page that says "We are passionate about quality" provides zero E-E-A-T signal. A page that says "Founded in 2018, 15,000 customers, workshop in Lyon, ISO 9001 certified" provides **verifiable facts** that AI can cross-reference with other sources.

### Accessible and complete policy pages

Shipping and return pages must be in native HTML and contain precise information: delivery times in days, return conditions, costs. An empty page or a JavaScript-only page invisible to crawlers is a negative E-E-A-T signal.

## Common mistakes that destroy your E-E-A-T signals

1. **Badges as images without alt text**: a Trustpilot badge displayed as `<img src="trustpilot-badge.png">` without alt text is invisible to AI crawlers. The fix: expose the data as JSON-LD `AggregateRating`.

2. **Reviews loaded only via JavaScript**: review widgets that load reviews via JavaScript after initial render are invisible to AI crawlers that read raw HTML. The fix: inject reviews via SSR (server-side rendering) or JSON-LD.

3. **Unverifiable claims**: "Best value for money" or "Category leader" without a source, study, or certification is an empty claim. AI agents treat it as marketing noise and penalize it.

4. **Missing or broken return policy**: if `/policies/refund-policy` returns a 404 error, the AI agent considers the return policy nonexistent, even if it's mentioned in the footer.

5. **Price mismatch between schema and display**: a price of 29.90 EUR displayed on the page but 39.90 EUR in the `Offer` schema creates a maximum distrust signal. AI may exclude the product from its recommendations entirely.

## Priority E-E-A-T checklist for AI

1. Implement `AggregateRating` in JSON-LD with `reviewCount` and `bestRating`
2. Expose certifications as `additionalProperty` in the Product schema
3. Write an "About Us" page with verifiable facts (founding date, customer count, location, certifications)
4. Verify that `/policies/shipping-policy` and `/policies/refund-policy` are accessible and complete
5. Remove marketing claims not backed by structured proof
6. Convert image-only badges to JSON-LD data
7. Verify coherence between displayed values and schema.org values
8. Publish expert content with verifiable sources

To assess your international readiness, also see our guide on [GEO Readiness](/en/kb/geo-readiness).

---

*Measure your trust signals: [start the free trial on the Shopify App Store](https://apps.shopify.com/verity-score).*

---

## Pricing and AI Visibility: How Agents Compare Your Products

## Why pricing matters more for AI than for humans

A human visitor compares 3 to 5 stores before buying. An AI agent like ChatGPT Shopping, Perplexity, or Google AI Mode compares hundreds of stores in seconds. Price is the first sorting criterion: if your price is unreadable or inconsistent, your product is eliminated before it is even evaluated on quality.

Since early 2026, AI shopping agents index prices in real time to build comparison tables. A product with properly structured pricing in [schema.org](/en/kb/schema-org) appears in these comparisons. A product whose price is only rendered via JavaScript or embedded in an image is simply invisible.

## How AI agents read your prices

AI crawlers follow a strict reading hierarchy to extract prices:

1. **Schema.org (Offer.price + priceCurrency)**: the most reliable source. AI agents read the JSON-LD in the page's `<head>`. If the price is there with the currency, the product is immediately comparable.
2. **Native HTML**: if schema.org is missing, agents look for the price in the HTML DOM (tags like `<span>`, `<p>` with classes such as `.price`, `.product-price`). Less reliable, but usable.
3. **JavaScript only**: if the price only appears after JavaScript execution (dynamic widget, third-party app), most AI crawlers cannot see it. Your product drops out of the comparison.

The rule is straightforward: if it is not in schema.org or static HTML, it does not exist for AI agents.

## The 5 pricing signals AI evaluates

### 1. Base price (Offer.price + priceCurrency)

The foundational signal. Every product page must contain in its JSON-LD:

```json
"offers": {
  "@type": "Offer",
  "price": "29.90",
  "priceCurrency": "EUR",
  "availability": "https://schema.org/InStock"
}
```

Without `price` and `priceCurrency`, the AI agent cannot compare your product. It is excluded from rankings.

### 2. Compare at price (original vs discounted)

On Shopify, the `compareAtPrice` represents the pre-discount price (the strikethrough price). When properly exposed, AI agents identify your product as being on sale and prioritize it in "best deals" recommendations.

The `compareAtPrice` must always be **higher** than the current price. If your current price is 29.90 EUR and your compareAtPrice is 24.90 EUR, that is an inconsistency that AI agents detect immediately  - and your credibility takes a hit.

### 3. Variant pricing

Modern AI agents evaluate prices per variant (size, color, volume). If your schema.org contains only a single product-level price while your variants have different prices, the agent displays a potentially incorrect price.

Each variant should have its own `Offer` block with its specific price. On Shopify, recent themes (Dawn, Refresh) automatically expose per-variant prices in JSON-LD. Older themes require a manual edit to the `main-product.liquid` template.

### 4. Currency consistency

AI agents cross-reference the currency declared in schema.org (`priceCurrency`) with the currency displayed on the page. If your schema.org says `EUR` but your page shows `$29.90`, the agent detects an inconsistency and reduces confidence in all of your structured data.

This issue is common on multi-currency stores that use conversion apps without syncing the schema.org. The currency must be identical across the JSON-LD, the displayed HTML, and the OG meta tags.

### 5. Price + availability = recommendable

An AI agent only recommends a product that is **both** in stock and has a readable price. Missing `availability: InStock` in your schema.org, even with a correct price, excludes your product from recommendations. Both signals must be present together.

## Common Shopify pricing mistakes

Here are the most frequent pricing errors that Verity Score detects on Shopify stores:

1. **Price only in a JavaScript widget**: some apps (currency converters, bundle builders) display the price via JS without updating the JSON-LD. Result: the price is invisible to AI crawlers.

2. **Inverted compareAtPrice**: the strikethrough price is lower than the current price (29.90 EUR crossed out for a current price of 34.90 EUR). AI agents interpret this as a price increase, not a promotion.

3. **Inconsistent multi-currency**: the store displays prices in CAD for Canadian visitors, but the schema.org stays in EUR. The AI agent sees two different prices for the same product and loses confidence.

4. **Missing priceCurrency**: the schema.org contains `"price": "29.90"` without `"priceCurrency": "EUR"`. Without a currency, the price is unusable for international comparison.

## Promotions and AI visibility

AI agents actively search for promotions to recommend the best offers. To make your promotions visible, structure them in schema.org:

- **Discounted price**: use `priceSpecification` with `price` (discounted price) and `referencePrice` (original price) in the Offer block.
- **Promotion dates**: add `priceValidUntil` so the agent knows when the offer expires. Promotions with an end date rank better because they create verifiable urgency.
- **Promo codes**: if your promotion requires a code, mention it in native HTML (not only in a JavaScript pop-up). AI agents like [ChatGPT Shopping](/en/kb/sell-on-chatgpt-shopify) and [Perplexity Shopping](/en/kb/perplexity-shopping) can read and pass promo codes to their users.

## Priority actions checklist

1. **Verify the JSON-LD** on every product page: `Offer.price`, `priceCurrency`, and `availability` must be present and correct
2. **Fix compareAtPrice**: it must always be higher than the current price (never inverted)
3. **Expose per-variant prices**: every variant with a different price needs its own Offer block
4. **Sync currencies**: the currency in schema.org, HTML, and OG meta must be identical
5. **Structure your promotions**: add `priceValidUntil` and `priceSpecification` for every active offer
6. **Test with a crawler**: disable JavaScript and verify that the price remains visible in the HTML source

## Next step

Pricing is a necessary but not sufficient signal. To maximize your AI visibility, combine well-structured pricing with a [complete schema.org](/en/kb/schema-org), clear shipping and return policies, and verifiable customer reviews.

---

*Audit your pricing visibility: [start the free trial on the Shopify App Store](https://apps.shopify.com/verity-score).*

---

## What is GEO (Generative Engine Optimization)? The Definitive Guide

## What is Generative Engine Optimization (GEO)?

**Generative Engine Optimization (GEO)** is the practice of optimizing websites  - particularly e-commerce stores  - to be cited, compared, and recommended by AI-powered search engines such as ChatGPT, Perplexity, and Google AI Mode. Unlike traditional SEO which optimizes for link-based rankings, GEO ensures your content and products appear in AI-generated responses. Where SEO asks "how do I rank on page one?", GEO asks "how do I get recommended when someone asks an AI for advice?"

In concrete terms: when a shopper types "best running shoes for flat feet under $150" into ChatGPT or Perplexity, GEO determines whether your product gets mentioned  - or whether your competitor does.

## Why GEO matters now

The shift from search-and-click to ask-and-receive is accelerating faster than most merchants realize.

### AI search is no longer experimental

ChatGPT processes hundreds of millions of queries weekly. Perplexity has become a daily tool for product research. Google AI Mode  - launched in 2025 and expanding rapidly in 2026  - now generates synthesized answers for a growing share of commercial queries. These are not niche tools anymore. They are becoming the default starting point for buying decisions.

### Commerce has entered the AI channel

ChatGPT Shopping launched with native product cards, price comparisons, and direct checkout links. Perplexity integrates merchant product feeds. Google's Shopping Graph powers AI-generated product recommendations. The infrastructure for [agentic commerce](/en/kb/what-is-agentic-commerce)  - where AI agents browse, compare, and buy on behalf of users  - is being built right now.

### The GEO market is exploding

The GEO industry is growing at a CAGR of 34%, from $886 million in 2024 to a projected $7.3 billion by 2031 (source: Incremys, 2026). Google AI Overviews now appear on **48% of all queries** - up from just 2.5% in early 2024. When an AI Overview appears, organic click-through rates drop 61% (Pew Research Center, 68,000 queries). Gartner projects that 25% of traditional organic search traffic will shift to AI chatbots and voice assistants by end of 2026. Fewer than 12% of marketing teams have a documented GEO strategy today.

### The cost of ignoring GEO is growing

Every day your store is invisible to AI systems, you lose potential revenue to competitors who have optimized for this channel. Unlike search optimization, where catching up means months of backlink building, GEO readiness can be improved in days with the right technical changes. But the window is closing: as more merchants optimize, the bar for AI visibility keeps rising.

## The 5 core principles of GEO

GEO rests on five foundational principles. Each one answers a specific question that AI systems ask when deciding whether to cite your content or recommend your products.

### 1. Citability  - Can AI extract precise facts?

AI engines cite sources that provide clear, unambiguous statements. A product page that says "lightweight and comfortable" gives the AI nothing to cite. A page that says "weighs 245g, EVA foam midsole, 8mm heel-to-toe drop" gives it three citable data points.

Citability means structuring your content so that every claim, specification, and feature is a discrete, extractable fact.

### 2. Comparability  - Can AI compare you fairly?

When a user asks "compare product A vs product B," the AI needs structured, standardized data. [Schema.org markup](/en/kb/schema-org), consistent units, and machine-readable specifications allow AI systems to build accurate comparison tables. Without structured data, the AI either skips you or guesses  - and guesses create hallucinations about your products.

### 3. Verifiability  - Are your claims backed by proof?

AI systems are increasingly trained to prefer verifiable claims over marketing assertions. A product page that states "rated 4.8 stars" needs an `AggregateRating` in its schema markup to prove it. A "free shipping" claim needs a visible, crawlable shipping policy page. The more your claims are machine-verifiable, the more AI systems trust  - and recommend  - your store.

### 4. Accessibility  - Can AI crawlers read your content?

If your content lives behind JavaScript widgets, rendered only client-side, or blocked by robots.txt rules that deny AI crawlers, it effectively does not exist for generative engines. GEO requires that your critical content  - product data, reviews, policies, specifications  - is accessible to [AI crawlers](/en/kb/robots-crawlers) in raw HTML or structured data formats.

### 5. Freshness  - Is your content current and dated?

AI systems weigh content freshness heavily. A product page last updated in 2023 carries less authority than one updated this month. Timestamps, recent reviews, current pricing, and dated policy pages signal that your information is reliable and current.

## GEO for e-commerce: what's different

E-commerce GEO has specific requirements that go beyond content optimization. AI shopping agents don't just read articles  - they evaluate products, compare prices, and verify purchase conditions.

**Product pages need schema.org Product and Offer markup.** This is non-negotiable. Without [JSON-LD Product schema](/en/kb/schema-org), AI systems cannot reliably extract your product name, price, availability, or variants. Every product page should include `Product`, `Offer`, and ideally `AggregateRating` markup.

**Reviews must be in HTML, not JavaScript widgets.** Many review apps (Yotpo, Judge.me, Loox) render reviews client-side via JavaScript. AI crawlers often cannot execute JS. Reviews must be server-side rendered (SSR) or included as structured data in JSON-LD to count for GEO.

**Prices must be machine-readable.** A price displayed only as an image, embedded in a button label, or formatted inconsistently across variants creates confusion for AI systems. Use `schema.org/Offer` with explicit `price` and `priceCurrency` fields.

**Policies must be accessible pages, not PDFs.** Return policies, shipping information, and warranty terms locked in PDF files are invisible to most AI crawlers. These should be standalone HTML pages, crawlable and linkable.

## GEO vs SEO: a quick comparison

| Dimension | SEO | GEO |
|-----------|-----|-----|
| **Optimizes for** | Search engine rankings (Google, Bing) | AI-generated responses (ChatGPT, Perplexity, AI Mode) |
| **Key signal** | Backlinks, keywords, domain authority | Structured data, citability, verifiability |
| **Content format** | Keyword-optimized pages | Machine-readable facts and schema markup |
| **User interaction** | Click on a blue link | Receive an AI-synthesized answer |
| **Measurement** | Rankings, organic traffic, CTR | AI citation rate, recommendation frequency, GEO Score |
| **Timeline to impact** | Months (backlink building) | Days to weeks (technical optimization) |

GEO does not replace SEO. The two disciplines are complementary. A strong SEO foundation  - quality content, fast pages, proper indexing  - supports GEO readiness. But SEO alone is no longer sufficient to capture AI-driven commerce. For a deeper comparison, see our full article on [GEO vs SEO](/en/kb/geo-vs-seo).

## The GEO tech stack

To be GEO-ready, your store needs a specific set of technical building blocks:

1. **robots.txt open to AI crawlers.** Ensure GPTBot (OpenAI), PerplexityBot, ClaudeBot, and Google-Extended are not blocked. Review your [robots.txt and crawler configuration](/en/kb/robots-crawlers).
2. **schema.org Product markup.** Every product page should have JSON-LD with `Product`, `Offer`, `brand`, `gtin`, `image`, and `description` at minimum.
3. **AggregateRating in JSON-LD.** If you have reviews, include `AggregateRating` with `ratingValue`, `reviewCount`, and `bestRating` in your schema.
4. **Agent discovery files.** Publish `/.well-known/agent-card.json` for store identity, capabilities and canonical citation targets. Use `llms.txt` as a supporting content index, not as a replacement for source pages.
5. **Server-side rendering (SSR).** Critical content  - reviews, specifications, prices, policies  - must be in the initial HTML response, not loaded via client-side JavaScript.
6. **Conversational content structure.** Structure your content as answers to questions buyers actually ask. Use clear headings, direct statements, and FAQ patterns that AI systems can extract and cite.

## How to measure GEO

Traditional web analytics do not capture AI visibility. You cannot see in Google Analytics whether ChatGPT recommended your product. GEO measurement requires a different approach.

A **GEO Score** evaluates your store's technical readiness across multiple dimensions: schema quality, crawler accessibility, content citability, review visibility, policy accessibility, and more. It answers a simple question: if an AI engine visited your store right now, how much useful data could it extract?

Verity Score provides a [GEO readiness audit](/en/kb/geo-readiness) that scans your store across 100+ signals and produces an actionable score with specific recommendations for improvement.

## Getting started: 5 first steps

If you are new to GEO, start here:

1. **Audit your robots.txt.** Check whether GPTBot, PerplexityBot, and ClaudeBot are allowed or blocked. If blocked, open access to your product and content pages.
2. **Add JSON-LD Product schema to every product page.** If you are on Shopify, your theme may already include basic schema  - but verify it includes `Offer`, `brand`, `gtin`, and `AggregateRating`.
3. **Check your reviews rendering.** View your product page source code (not the rendered page). If reviews are absent from the HTML, they are invisible to AI. Ask your review app provider about SSR support.
4. **Create standalone policy pages.** Ensure your shipping, return, and warranty policies are accessible HTML pages linked from your navigation  - not buried in PDFs or accordion widgets.
5. **Run a GEO audit.** Get a baseline score so you know where you stand and what to prioritize.

GEO is not a future concern. It is a present reality shaping how products are discovered, compared, and purchased. The merchants who optimize now will capture the AI commerce wave. Those who wait will wonder where their traffic went.

---

*Measure your GEO readiness: [start the free trial on the Shopify App Store](https://apps.shopify.com/verity-score).*

---

## How to Sell on ChatGPT Shopping: Step-by-Step for Shopify (2026)

## What is ChatGPT Shopping?

ChatGPT Shopping is OpenAI's product discovery and purchase feature built into ChatGPT. It allows users to search, compare, and buy products directly in the chat interface. For Shopify merchants, it represents a new zero-ad-cost acquisition channel  - but only if your store is optimized for AI visibility.

Since early 2026, ChatGPT has been surfacing product cards with prices, images, ratings, and direct checkout links inside conversational answers. Unlike Google Shopping, there is no bidding. Unlike social ads, there is no creative fatigue. The products that appear are the ones ChatGPT can read, trust, and recommend.

## How ChatGPT Shopping works (merchant perspective)

The flow is straightforward:

1. **A user asks a product question**  - "What's the best moisturizer for dry skin under $40?" or "I need a travel backpack with a laptop compartment."
2. **ChatGPT crawls and compares available stores**  - it reads product pages, structured data, reviews, and policies from stores it can access.
3. **Product cards are displayed**  - each card includes the product name, price, image, star rating, and store name.
4. **Checkout happens via ACP or redirect**  - the user either completes the purchase through the [Agentic Commerce Protocol (ACP)](/en/kb/acp-agentic-commerce-protocol) or is redirected to the merchant's checkout page.

The critical point: if ChatGPT cannot crawl your store, read your structured data, or verify your policies, your products will not appear. There are no ads to buy your way in.

## The 3 layers of ChatGPT visibility

Getting your products into ChatGPT Shopping requires three distinct layers. Missing any one of them reduces or eliminates your chances.

### Layer 1: Discoverability

ChatGPT uses GPTBot to crawl stores. If your [robots.txt blocks GPTBot](/en/kb/robots-crawlers), your store is invisible.

**What to do:**
- Verify that `robots.txt` allows GPTBot (User-agent: GPTBot / Allow: /)
- Create a `llms.txt` file at your domain root listing your key pages  - product categories, bestsellers, policy pages
- Ensure your sitemap.xml is accessible and up to date

### Layer 2: Readability

Once GPTBot can access your pages, ChatGPT needs to understand them. This means machine-readable [structured data](/en/kb/schema-org).

**What to do:**
- Implement complete `schema.org/Product` markup on every product detail page (PDP)
- Include `Offer` with price, currency, and availability in native HTML  - not rendered by JavaScript only
- Add [AggregateRating](/en/kb/aggregate-rating) to expose your review data (star rating, review count) in JSON-LD
- Make sure product descriptions, specs, and ingredients are in the HTML source, not loaded asynchronously

### Layer 3: Transactability

ChatGPT Shopping prioritizes stores where the purchase can be completed smoothly. This is where the [Agentic Commerce Protocol (ACP)](/en/kb/acp-agentic-commerce-protocol) comes in.

**What to do:**
- Ensure your checkout flow works without JavaScript-dependent popups or multi-step redirects
- Publish clear, crawlable shipping and return policies as standalone HTML pages (not PDFs)
- Expose `shippingDetails` and `hasMerchantReturnPolicy` in your Product schema
- If ACP-compliant checkout is available for your platform, enable it

## Step-by-step setup for Shopify

Here is the practical checklist for [Shopify merchants who want to sell on ChatGPT](/en/kb/sell-on-chatgpt-shopify):

1. **Allow GPTBot in robots.txt**  - Edit your `robots.txt.liquid` file in your Shopify theme to include `User-agent: GPTBot` with `Allow: /`. Many Shopify themes block AI crawlers by default since late 2025.

2. **Complete schema.org Product on every PDP**  - Most Shopify themes include basic Product + Offer schema. Verify that `brand`, `sku`, `gtin`, `description`, `image`, and `availability` fields are populated. Use Google's Rich Results Test or a JSON-LD validator.

3. **Expose reviews as AggregateRating in JSON-LD**  - If you use Judge.me, Loox, Yotpo, or Stamped, check whether the app injects [AggregateRating](/en/kb/aggregate-rating) into the page's JSON-LD. Many review apps render stars visually but do not add structured data. Add it manually in your theme's `product.liquid` if needed.

4. **Create llms.txt with key pages**  - Add a `llms.txt` file at your store root listing: homepage, top collections, bestseller PDPs, shipping policy, return policy, FAQ, and about page.

5. **Ensure native HTML content**  - Shopify's Liquid templates render server-side, which is good. But custom sections using JavaScript-only rendering (React widgets, lazy-loaded tabs) will be invisible to GPTBot. Critical product information  - price, description, specs, reviews  - must be in the initial HTML response.

6. **Write clear policies as accessible pages**  - Shipping, returns, warranty, and privacy policies should be standalone pages (`/policies/shipping-policy`, `/policies/refund-policy`), not embedded in PDFs or hidden behind accordions that require JavaScript to open.

## ACP (Agentic Commerce Protocol) explained

The Agentic Commerce Protocol is an emerging standard that allows AI agents  - including ChatGPT  - to complete purchases on behalf of users. Instead of redirecting to a checkout page, ACP enables the agent to add items to cart, apply discounts, and process payment within the chat interface.

For Shopify merchants, ACP support is still rolling out. The key preparation steps are: clean checkout flows, machine-readable policies, and complete product structured data. For a deeper dive, read our [full ACP guide](/en/kb/acp-agentic-commerce-protocol).

## Common mistakes that block ChatGPT recommendations

These are the errors we see most frequently when auditing Shopify stores for AI visibility:

| Mistake | Why it matters |
|---------|---------------|
| GPTBot blocked in robots.txt | ChatGPT cannot crawl any page on your store |
| Prices rendered in JavaScript only | GPTBot sees an empty price field; no product card can be generated |
| Review widget without AggregateRating | Stars appear visually but ChatGPT cannot read the rating data |
| Policy pages published as PDFs | AI crawlers cannot reliably extract text from PDF files |
| Inconsistent pricing between schema and page | Price mismatch triggers a trust penalty; ChatGPT may skip the product |
| Missing product images in schema | Product cards without images get significantly lower engagement |

## Measuring your ChatGPT readiness (GEO Score)

Your GEO (Generative Engine Optimization) score measures how well your store is prepared for AI-driven commerce channels like ChatGPT Shopping, Perplexity Shopping, and Google AI Mode.

A GEO audit checks all three layers  - discoverability, readability, and transactability  - and produces an actionable report with specific fixes for your Shopify store.

The stores that rank highest in ChatGPT Shopping are the ones that treat AI visibility as a technical discipline, not an afterthought.

*Check your ChatGPT Shopping readiness: [start the free trial on the Shopify App Store](https://apps.shopify.com/verity-score).*

---

## State of AI Commerce 2026: ChatGPT Shopping, AI Mode & Agentic Storefronts

## The AI commerce revolution is here

The shift from hypothetical to operational happened faster than most merchants expected. In 2026, AI-powered shopping experiences handle a meaningful share of product discovery and purchase decisions. Consumers ask ChatGPT to find running shoes, let Google AI Mode compare kitchen appliances, and rely on Perplexity to shortlist supplements based on specific health goals.

This is not a future scenario. It is the current state of e-commerce. And for Shopify merchants, the question is no longer whether AI commerce matters, but whether their store is visible to the systems driving it.

## The 4 AI commerce platforms reshaping e-commerce

Four major platforms now mediate the relationship between shoppers and products. Each operates differently, but all share one requirement: they need structured, crawlable, verifiable product data to work.

### ChatGPT Shopping (OpenAI)

OpenAI's shopping experience lets users discover products through natural conversation. A shopper types "waterproof hiking boots under $150 with good ankle support" and receives curated recommendations with images, prices, and reviews. Since early 2026, ChatGPT supports native checkout through the [Agentic Commerce Protocol (ACP)](/en/kb/acp-agentic-commerce-protocol), powered by Stripe. Products that surface here benefit from high purchase intent and minimal comparison friction.

### Google AI Mode (Gemini)

Google's AI Mode integrates directly with the Shopping Graph, a structured product database containing billions of listings. When a user asks Gemini a product question, the response draws on schema.org data, merchant feeds, and real-time inventory signals. Stores with complete structured data and active Merchant Center feeds get prioritized. Google backs the [Universal Commerce Protocol (UCP)](/en/kb/ucp-universal-commerce-protocol), an open standard designed to let any AI agent transact with any merchant.

### Perplexity Shopping

Perplexity approaches product discovery as an answer engine. When users ask product-related questions, Perplexity synthesizes information from crawled pages, reviews, and structured data to generate recommendations with source citations. Stores with rich, crawlable HTML content, visible reviews in the source code, and clear product specifications gain a significant advantage here. Perplexity launched **Snap to Shop** in February 2026 (visual search from photos), operates without ads entirely (100% subscription revenue model, $500M ARR target), and the 9th Circuit paused Amazon's Comet injunction in March 2026, keeping Perplexity's browser agent live during appeal.

### Autonomous shopping agents

The newest and most disruptive category. AI agents like OpenAI Operator and emerging multi-agent systems browse the web, compare products across stores, and can complete purchases on behalf of users. These agents evaluate stores on technical criteria: is the price machine-readable? Are reviews in the HTML source? Is checkout automatable? The [agentic commerce](/en/kb/what-is-agentic-commerce) paradigm rewards stores that expose clean, structured signals.

## The protocol wars: ACP vs UCP

Two competing protocols are defining how AI agents transact with merchants. Understanding both is essential for any store preparing for AI commerce.

| | **ACP (Agentic Commerce Protocol)** | **UCP (Universal Commerce Protocol)** |
|---|---|---|
| **Backed by** | OpenAI + Stripe | Google + Shopify + Walmart + Etsy |
| **Approach** | Native checkout within ChatGPT | Open standard, any AI agent can use |
| **Checkout** | Redirect to merchant site (post-March 2026 pivot) | Merchant-hosted, agent-initiated |
| **Integration** | Stripe merchant account required | `.well-known/ucp` manifest + Shopping Graph |
| **Scope** | ChatGPT ecosystem | Cross-platform, cross-agent |
| **Status (April 2026)** | Visual discovery live, checkout on merchant site | Cart + Catalog + Identity Linking update (March 2026) |

The strategic reality: merchants cannot afford to pick one. ChatGPT Shopping drives high-intent conversational traffic through ACP. Google AI Mode and the broader agent ecosystem rely on UCP. Preparing for both protocols is not optional; it is the baseline for [AI commerce readiness](/en/kb/geo-readiness).

## What the data tells us

Verity Score audits hundreds of Shopify stores for AI visibility. The results reveal a significant gap between what AI systems need and what most stores provide.

**Most Shopify stores block at least one major AI crawler.** Default robots.txt configurations on many themes and apps still block GPTBot, Google-Extended, or PerplexityBot. A store that is invisible to AI crawlers cannot appear in AI-powered recommendations, regardless of product quality.

**Schema.org completeness averages only 40-60%.** Most stores have basic Product schema, but critical fields like `brand`, `AggregateRating`, `shippingDetails`, `returnPolicy`, and `hasVariant` are often missing or incomplete. AI agents rely heavily on these fields for cross-store comparison.

**Most stores still lack dedicated AI discovery surfaces.** A useful setup combines `/.well-known/agent-card.json` for identity, capabilities and canonical citation targets with `llms.txt` as a supporting content index. These files help orientation and tracking, but the primary citation layer remains crawlable HTML, schema.org, product feeds, reviews, policies, sitemap coverage and internal links.

**AggregateRating in HTML is missing on the majority of stores.** Review apps like Loox, Judge.me, and Yotpo typically load ratings via JavaScript. The result: the HTML source that AI crawlers read contains zero review data, making the store appear unreviewed.

These findings represent patterns across hundreds of audits. For the full benchmark data and methodology, see the [State of AI Commerce report](/en/state-of-ai-commerce).

## 3 trends to watch in 2026-2027

### AI-first checkout

Autonomous agents are beginning to complete purchases without human interaction. A user sets preferences, budget constraints, and reorder rules. The agent monitors prices, evaluates options, and executes the purchase. Stores that support automated checkout flows through ACP or UCP will capture this emerging transaction channel.

### Multi-agent comparison

Multiple AI agents evaluating the same store simultaneously is becoming common. A shopper might ask ChatGPT, Perplexity, and Google AI Mode the same question. Each agent crawls the store independently. Inconsistencies between what each agent extracts (different prices, missing reviews on one platform, incomplete specs on another) erode trust and reduce recommendations.

### GEO as a standard marketing discipline

Generative Engine Optimization is following the trajectory that SEO took in the 2010s. What started as a niche technical concern is becoming a standard marketing discipline. By 2027, every serious e-commerce brand will have a GEO strategy alongside their paid and organic strategies. The merchants who invest now will have a measurable advantage when the market matures.

## What merchants should do now

Five concrete steps to prepare your Shopify store for AI commerce:

1. **Audit your AI visibility (GEO Score).** Before optimizing, measure. A GEO audit reveals how AI systems currently see your store: what they can crawl, what structured data they find, and where the gaps are.

2. **Open robots.txt to AI crawlers.** Ensure GPTBot, Google-Extended, PerplexityBot, ClaudeBot, and other AI user agents are not blocked. This is the single highest-impact, lowest-effort change a merchant can make.

3. **Complete schema.org Product markup.** Go beyond basic Product schema. Include `brand`, `AggregateRating` (rendered in HTML, not just JavaScript), `shippingDetails`, `returnPolicy`, `hasVariant` with individual variant pricing, and `Offer` with `availability`. See the [Shopify schema.org guide](/en/blog/shopify-schema-org-guide) for implementation details.

4. **Publish agent discovery surfaces.** Add `/.well-known/agent-card.json` with your store identity, supported capabilities and canonical URLs. Add `llms.txt` as a concise supporting index to product categories, policies, proof pages and best sellers.

5. **Prepare for both ACP and UCP.** If you use Stripe, explore ACP merchant integration for ChatGPT Shopping. Publish a `.well-known/ucp` manifest to signal agent readiness. Neither protocol requires a complete overhaul; both start with structured data and accessibility.

## The bottom line

AI commerce is not a future disruption. It is an active, growing channel. The stores that AI agents recommend today are the ones with clean structured data, open crawling policies, and protocol readiness. The stores they skip are invisible, no matter how good their products are.

The merchants who optimize for AI visibility now will capture first-mover advantage in a channel that is growing every quarter. Those who wait will find themselves competing for attention in a marketplace where the rules have already been set.

*Where does your store stand? Find out: [start the free trial on the Shopify App Store](https://apps.shopify.com/verity-score).*

---


# Base de connaissances  - Français

## Qu'est-ce que le commerce agentique ? Guide complet 2026

Le commerce agentique transforme l'e-commerce : des agents IA autonomes recherchent, comparent et achètent pour les consommateurs. Ce guide couvre les protocoles ACP et UCP, et explique comment préparer votre store Shopify pour ce nouveau paradigme.

URL: /fr/kb/commerce-agentique

---

## GEO SEO : GEO vs SEO, les différences en e-commerce

Le SEO optimise pour les liens bleus. Le GEO optimise pour les réponses IA. Ce guide compare les deux disciplines, explique leurs complémentarités et pourquoi le GEO est devenu indispensable pour les marchands Shopify en 2026.

URL: /fr/kb/geo-vs-seo

---

## Comprendre votre score GEO : 9 facteurs décryptés

Le GEO Score mesure la visibilité de votre store Shopify pour les moteurs de recherche génératifs et les agents IA. Découvrez les 9 facteurs et comment les optimiser pour maximiser votre recommandabilité.

URL: /fr/kb/geo-readiness

---

## Audit GEO : le guide complet pour optimiser votre store Shopify pour l'IA

Un audit GEO mesure la visibilité de votre store Shopify dans ChatGPT, Perplexity et Google AI Mode. Guide complet avec les 4 piliers, 9 facteurs et une checklist actionnable.

URL: /fr/kb/audit-geo

---

## Vendre dans ChatGPT : le guide Shopify complet pour 2026

Comment vendre vos produits Shopify via ChatGPT Shopping et les Agentic Storefronts. Couvre les 8 critères de visibilité, une checklist actionnable et un cas concret de progression de 35 à 78/100.

URL: /fr/kb/vendre-chatgpt-shopify

---

## ACP  - Agentic Commerce Protocol : guide OpenAI pour marchands Shopify

L'ACP (Agentic Commerce Protocol) est une spécification de commerce publiée par OpenAI. La documentation actuelle distingue clairement ce qui est ouvert à tous de ce qui reste propre aux partenaires approuvés : construire avec l'ACP est ouvert, tandis que l'onboarding des flux produit et certains parcours Instant Checkout sont soumis à approbation. Shopify Catalog est documenté comme voie de diffusion des données produit. Ce guide ne présente pas 4 % comme une grille tarifaire officielle actuelle et sépare les prérequis vérifiables des programmes auxquels un marchand ne peut pas s'inscrire seul.

URL: /fr/kb/acp-agentic-commerce-protocol

---

## UCP  - Universal Commerce Protocol : guide complet pour marchands Shopify

Le UCP (Universal Commerce Protocol) est le standard co-développé par Google et Shopify. Ce guide couvre l'activation sur Shopify, les capabilities du protocole et l'audit de conformité UCP.

URL: /fr/kb/ucp-universal-commerce-protocol

---

## Perplexity Shopping : guide marchand Shopify fondé sur les sources officielles

Ce guide décrit les actions vérifiables dans Shopify : garder les données produit, prix et disponibilité à jour, publier des descriptions et FAQ utiles, rendre les avis accessibles et connecter les canaux marchands proposés par Shopify. Perplexity ne publie pas de formule de classement ni de promesse d'apparition : le guide sépare donc les prérequis documentés des bonnes pratiques et des résultats non garantis.

URL: /fr/kb/perplexity-shopping

---

## Schema.org Product : pourquoi et comment sur Shopify

Guide du balisage produit Shopify : exigences propres aux fiches marchandes Google, cohérence des valeurs et validation. Aucune garantie de recommandation IA. Voir le guide complet : [Schema.org Product sur Shopify](https://verityscore.io/fr/kb/schema-org/).

URL: /fr/kb/schema-org

---

## AggregateRating : rendre vos avis lisibles par l'IA

Les avis clients sont souvent invisibles pour les IA car chargés en JavaScript. Ce guide explique comment exposer votre AggregateRating dans le HTML de votre store Shopify pour les crawlers IA.

URL: /fr/kb/aggregate-rating

---

## robots.txt et crawlers IA : ne bloquez pas vos ventes

Comment le robots.txt affecte la visibilité de votre store Shopify pour les agents IA. Guide pratique pour ouvrir l'accès à GPTBot, OAI-SearchBot (citation ChatGPT Shopping), PerplexityBot et ClaudeBot sans compromettre la sécurité.

URL: /fr/kb/robots-crawlers

---

## llms.txt : index de contenu IA utile, pas signal principal

Le fichier llms.txt guide les LLM vers vos contenus prioritaires. Ce guide explique pourquoi l'ajouter à votre store Shopify et comment le rédiger pour orienter les agents IA vers vos pages clés.

URL: /fr/kb/llms-txt

---

## Prix et visibilité IA : comment les agents comparent vos produits

Les agents IA comparent les prix avant de recommander. Ce guide explique comment exposer vos prix, promotions et compareAtPrice pour maximiser votre visibilité dans ChatGPT Shopping, Perplexity et Google AI Mode.

URL: /fr/kb/prix-visibilite-ia

---

## Signaux E-E-A-T pour l'IA : comment les agents évaluent la confiance

Les agents IA évaluent votre store selon les critères E-E-A-T (expérience, expertise, autorité, fiabilité). Ce guide détaille les signaux de confiance qui déterminent si l'IA vous recommande ou vous ignore.

URL: /fr/kb/eeat-signaux-ia

---

## Claims & Proof : la crédibilité aux yeux de l'IA

Détection des claims marketing, vérification des preuves, analyse de la cohérence. Comment rendre vos promesses commerciales vérifiables pour les agents IA et éviter les signaux de méfiance.

URL: /fr/kb/claims-proof

---

## Contenu conversationnel : écrire pour les humains ET les IA

Comment structurer le contenu de vos fiches produit Shopify pour qu'il soit à la fois engageant pour les humains et exploitable par les agents IA en format question-réponse.

URL: /fr/kb/contenu-conversationnel

---

## Livraison et retours : signaux de confiance critiques

Les politiques de livraison et de retour sont des signaux de confiance majeurs pour les agents IA. Ce guide explique comment les rendre visibles et vérifiables sur Shopify avec schema.org.

URL: /fr/kb/livraison-retours

---

## AI Buyer Score : la checklist des agents d'achat

Comment un agent d'achat IA évalue votre store Shopify. Les 9 critères de recommandabilité et les actions concrètes pour passer de "non recommandé" à "recommandé".

URL: /fr/kb/ai-buyer-score

---


# Blog  - Français

## llms.txt en 2026 : utile, mais pas magique

Un guide de décision sourcé pour les équipes e-commerce. Il explique ce que change llms.txt v2, les positions documentées de Google, OpenAI et Anthropic, et pourquoi une publication ne prouve ni récupération, ni citation, ni revenu. Il propose un contrôle Shopify avant toute création, un test qualité en huit points, une matrice de priorité et une frontière claire entre carte de contenu et protocoles transactionnels comme ACP et UCP.

URL: /fr/blog/llms-txt-2026-ecommerce
Markdown: /fr/blog/llms-txt-2026-ecommerce.md
EN URL: /en/blog/llms-txt-2026-ecommerce
EN Markdown: /en/blog/llms-txt-2026-ecommerce.md

---

## AI Max Shopping : données produit Shopify et IA

Google a annoncé le 30 avril 2026 l'arrivée d'AI Max dans les campagnes Shopping. L'article explique pourquoi ce signal dépasse Google Ads : AI Max for Shopping utilise les flux Merchant Center pour transformer les données produit en annonces Shopping dynamiques capables de répondre à des requêtes conversationnelles, tandis que ChatGPT Shopping et les Agentic Storefronts de Shopify confirment la même tendance.

Doctrine centrale : une fiche produit Shopify devient un contrat de données entre Shopify Admin, Merchant Center, HTML source, HTML rendu, JSON-LD Product/Offer/AggregateRating, métadonnées, avis, politiques et fichiers publics de découverte. L'article sépare les `productRecommendations` locales des `globalBlockers` site-wide, et définit le workflow produit : diagnostic, patch preview, vérification post-fix. Trois diagrammes SVG expliquent le contrat de données produit, la boucle d'audit granulaire et la séparation productRecommendations/globalBlockers.

Contrôles prioritaires :
1. Corriger les contradictions prix et stock entre Shopify Admin, HTML, JSON-LD et feed.
2. Rendre les avis visibles aussi lisibles par machine via AggregateRating.
3. Compléter Product et Offer : marque, SKU, GTIN, variantes, disponibilité, priceCurrency, shippingDetails et returnPolicy.
4. Transformer les claims marketing en preuves vérifiables.
5. Ajouter du contenu conversationnel répondant aux questions naturelles des acheteurs.
6. Utiliser le HTML rendu quand une donnée produit critique est générée par JavaScript, apps ou widgets.
7. Aligner Merchant Center uniquement pour le produit vérifié quand il diverge.

URL: /fr/blog/ai-max-shopping-donnees-produit-shopify
Markdown: /fr/blog/ai-max-shopping-donnees-produit-shopify.md
EN URL: /en/blog/ai-max-shopping-shopify-product-data
EN Markdown: /en/blog/ai-max-shopping-shopify-product-data.md

---

## GEO pour les marques beauté sur Shopify : le guide 2026

La beauté est la première catégorie où la recherche IA change les règles, parce que les assistants passent par les ingrédients avant la marque. Le 5 juin 2026, Kendo Brands a déclaré que le trafic ChatGPT vers Fenty Beauty a été multiplié par deux à trois en 2026 (Chief Marketer). L'adoption est rapide : 38% des acheteurs beauté US ont utilisé l'IA en 2026, contre 2% un an plus tôt (Tinuiti, source US). Le trafic référé par l'IA a converti 42% mieux que le non-IA en mars 2026 (Adobe, source US). Contrepoids honnête : seules 2,1% des conversations ChatGPT concernent des produits achetables (NBER). Statut France au 23 juillet 2026 : Google a déployé les Aperçus IA et le Mode IA en France le 22 juillet 2026, les surfaces utiles sont ChatGPT, Perplexity, Gemini et Mistral Le Chat.

Comment l'IA recommande un produit beauté : elle fait correspondre un nom INCI connu et sa concentration à une problématique de peau, filtre par type de peau et contre-indications, puis pèse les allégations étayées, certifications et avis tiers. Environ 90% des prompts beauté sont « meilleur X pour [type de peau/problématique] », et les moteurs ne sont pas d'accord entre eux (tester ChatGPT, Perplexity, Gemini, Mistral séparément).

Les 7 leviers on-page : (1) rendre INCI et concentrations lisibles par machine dans le titre, un tableau d'actifs et le additionalProperty du JSON-LD ; (2) baliser type de peau, problématique et contre-indications ; (3) étayer les allégations (convergence réglementaire-confiance IA : ce que le régulateur exige que vous prouviez, l'IA exige que vous le démontriez ; UE 655/2013, DGCCRF 46% d'allégations non conformes, ASA a confirmé une plainte contre Garnier « cliniquement prouvé en 2 semaines » en avril 2026 ; autorisé « aide à réduire l'apparence de » vs interdit « traite l'acné / stimule le collagène ») ; (4) contenu comparatif honnête et « pour qui ce n'est pas » ; (5) certifications comme jetons d'autorité lisibles par machine (Leaping Bunny, EWG Verified, COSMOS) ; (6) formule titre et description réponse-d'abord ; (7) schema FAQPage répondant aux vraies questions. Prérequis : avis rendus côté serveur et AggregateRating sur le Produit (pas l'Organisation, considérée comme auto-attribuée par Google).

Corrections couche technique : les données produit Shopify sont auto-intégrées à ChatGPT ; OpenAI recommande un snapshot complet quotidien via file upload plus des mises à jour intra-journée via l'API, formats Parquet/JSONL/CSV/TSV (pas XML), GTIN optionnel chez OpenAI ; résultats non publicitaires ; OpenAI s'éloigne du paiement instantané dans ChatGPT au profit du checkout marchand (juin 2026). robots.txt doit autoriser OAI-SearchBot, ChatGPT-User, PerplexityBot, Googlebot ; GPTBot et Google-Extended sont entraînement uniquement. Pas de type schema cosmétique, utiliser additionalProperty sur le Produit. Programme Marchand Perplexity gratuit pour les boutiques Shopify livrant aux US. Hors-site : Reddit influence surtout via les données d'entraînement, sélections éditoriales et avis retailers comptent, éviter l'astroturfing. Plan 30/60/90 inclus. Verity Score audite la verticale beauté : présence INCI et type de peau, allégations non vérifiables, avis JavaScript-only, éligibilité AggregateRating, accès crawlers IA, complétude de la fiche produit.

URL: /fr/blog/geo-beaute-cosmetiques-shopify
Markdown: /fr/blog/geo-beaute-cosmetiques-shopify.md
EN URL: /en/blog/geo-beauty-skincare-shopify
EN Markdown: /en/blog/geo-beauty-skincare-shopify.md

---

## Qu'est-ce que le GEO (Generative Engine Optimization) ? Le guide définitif

Le GEO est la discipline qui optimise votre site pour les moteurs de recherche génératifs. ChatGPT, Perplexity, Google AI Mode : ce guide couvre la définition, les 5 principes fondamentaux, la comparaison GEO vs SEO et la stack technique pour ne pas rater le virage IA du e-commerce.

URL: /fr/blog/quest-ce-que-le-geo

---

## Comment vendre sur ChatGPT Shopping : le guide marchand 2026

ChatGPT Shopping est un nouveau canal d'acquisition pour les marchands Shopify. Ce guide détaille comment apparaître dans les recommandations produit de ChatGPT, configurer l'ACP et optimiser votre visibilité IA.

URL: /fr/blog/vendre-chatgpt-shopping-guide

---

## L'état du commerce IA en 2026 : ce que les marchands doivent savoir

ChatGPT Shopping, Google AI Mode, Perplexity, agents autonomes : l'IA transforme le e-commerce. État des lieux des tendances, protocoles ACP vs UCP et opportunités concrètes pour les marchands Shopify en 2026.

URL: /fr/blog/etat-commerce-ia-2026

---

## Pourquoi le GEO est le nouveau CRO

Le CRO optimise le clic humain. Le GEO optimise la recommandation IA. Ce guide explique pourquoi ces deux disciplines sont complémentaires et pourquoi ignorer le GEO coûte déjà des ventes aux marchands Shopify.

URL: /fr/blog/pourquoi-geo-cro

---

## AI Buyer Score : votre store est-il recommandable par une IA ?

Comment un agent d'achat IA décide de recommander (ou pas) un produit Shopify. Ce guide détaille les 9 critères du AI Buyer Score et les actions concrètes pour optimiser chaque critère.

URL: /fr/blog/ai-buyer-score-explique

---

## Le guide Schema.org pour Shopify : ce que les IA lisent vraiment

Tutoriel pratique pour implémenter Product, Offer, AggregateRating et FAQ schema sur Shopify. Couvre ce que les thèmes font par défaut, ce qu'il faut ajouter et les erreurs courantes à éviter.

URL: /fr/blog/shopify-schema-org-guide

---

## HTML and Markdown Alternates

Strategic IA-readable pages expose synchronized Markdown alternates. Commercial pages, blog, KB, industry and glossary alternates are synchronized with their canonical HTML sources. The `.md` versions are intended for compact LLM ingestion and are tracked server-side as Markdown alternate or Accept:text/markdown access.

Coverage:
- Commercial SEO/GEO/AEO pages and comparatifs
- Blog index and every published blog article
- Knowledge Base index and every published KB guide
- Industry/sector guide index and every vertical guide
- GEO glossary FR/EN
- Legal/privacy trust and compliance pages FR/EN

Excluded from Markdown mirrors: admin, audit app shell, search, contact and operational pages. Legal/privacy pages are mirrored as trust/compliance documents and remain `noindex` in HTML. Build guard: `scripts/check-markdown-alternates.mjs`.

### Shopify GEO App

Public Shopify App Store listing: https://apps.shopify.com/verity-score

Verity Score includes a free plan, an AI Buyer Score measured across 8 dimensions, and continuous GEO auditing of Product schema, AggregateRating, Offers, variants, Shopify Markets consistency, policies, claims, robots.txt, sitemap.xml, agent-card, llms.txt and ai.txt.

HTML: /en/shopify-geo-app/
Markdown: /en/shopify-geo-app.md
FR HTML: /fr/app-shopify-geo/
FR Markdown: /fr/app-shopify-geo.md

### Product Overview

The six surfaces of the Shopify app presented as one loop: see what the assistants answer, find what caused it, fix the source, prove it moved. Covers the 9 GEO factors, 22 anti-hallucination detectors and 8 AI Buyer Score dimensions, and states plainly what the product does not do.

HTML: /en/product/
Markdown: /en/product.md
FR HTML: /fr/produit/
FR Markdown: /fr/produit.md

### Prompt Tracking (AI Shopper Tests)

Buyer decision questions run on ChatGPT, Perplexity, Gemini and Claude in the merchant market and language. Each answer is stored verbatim with its model, timestamp and source URLs, then classified as recommends you, names another shop, ignores you, or advises against. Rival shops named inside the same answers are extracted and counted.

HTML: /en/product/ai-shopper-tests/
Markdown: /en/product/ai-shopper-tests.md
FR HTML: /fr/produit/simulation-acheteurs/
FR Markdown: /fr/produit/simulation-acheteurs.md

### Product Answer Coverage

The 8 AI Buyer Score dimensions checked on every product, separating questions the merchant own published evidence answers today from questions waiting on merchant data. Missing facts are named in plain language and published back as public product facts once approved.

HTML: /en/product/product-answers/
Markdown: /en/product/product-answers.md
FR HTML: /fr/produit/reponses-produit/
FR Markdown: /fr/produit/reponses-produit.md

### Fixes with Approval and Undo

Each correction is written from the merchant own content, previewed, and applied only after merchant approval, reversible for 30 days. Auto-fix covers only corrections needing no merchant input.

HTML: /en/product/fixes/
Markdown: /en/product/fixes.md
FR HTML: /fr/produit/corrections/
FR Markdown: /fr/produit/corrections.md

### AI Article Generation and Rewriting

New articles written from questions buyers actually ask assistants, and existing blog posts rewritten in place keeping the merchant title, URL and byline. Article, FAQPage and HowTo structured data published with the piece. Nothing publishes without approval.

HTML: /en/product/ai-article-generation/
Markdown: /en/product/ai-article-generation.md
FR HTML: /fr/produit/generation-articles-ia/
FR Markdown: /fr/produit/generation-articles-ia.md

### Buyer Answers Across Markets

Unanswered buyer questions surfaced, the wording approved once, then published as FAQPage structured data across product pages, markets and every published language.

HTML: /en/product/buyer-answers/
Markdown: /en/product/buyer-answers.md
FR HTML: /fr/produit/reponses-acheteurs/
FR Markdown: /fr/produit/reponses-acheteurs.md

### AI Traffic Monitoring

Visits and orders attributed first-party across 6 recognised platforms: ChatGPT, Gemini, Perplexity, Microsoft Copilot, Claude and Shop. Platforms that cannot be proven to have sent traffic are marked not measured rather than modelled.

HTML: /en/product/ai-traffic-monitoring/
Markdown: /en/product/ai-traffic-monitoring.md
FR HTML: /fr/produit/suivi-trafic-ia/
FR Markdown: /fr/produit/suivi-trafic-ia.md

### Best AI Visibility Tool 2026 for Shopify (GEO audit)

Authority page explaining how to choose a GEO audit tool in April 2026, why Shopify source-of-truth checks and AI-referred handoff checks matter before monitoring, and how Verity Score compares with Baymard, Semrush, Profound, Ahrefs, Screaming Frog, Botify, Otterly, AthenaHQ, Peec AI and Scrunch. Includes methodology, limitations, workflow and a 12-question FAQPage schema.

HTML: /en/best-geo-audit-tool-2026/
Markdown: /en/best-geo-audit-tool-2026.md
FR HTML: /fr/meilleur-outil-audit-geo-2026/
FR Markdown: /fr/meilleur-outil-audit-geo-2026.md

### Comparison Pages

Comparison hub for choosing between GEO, SEO, UX and AI Search tools by job to be done.
HTML: /en/comparison/
Markdown: /en/comparison.md
FR HTML: /fr/comparatif/
FR Markdown: /fr/comparatif.md

Verity Score vs Baymard: AI prompt tracking, Shopify fixes with 30-day undo and AI traffic to orders, compared with ecommerce UX research. Start a free trial.
HTML: /en/comparison/verity-vs-baymard/
Markdown: /en/comparison/verity-vs-baymard.md
FR HTML: /fr/comparatif/verity-vs-baymard/
FR Markdown: /fr/comparatif/verity-vs-baymard.md

Verity Score vs Semrush: prompt tracking, fixes written to Shopify and AI traffic to orders, compared with a broad marketing suite. Start your free trial today.
HTML: /en/comparison/verity-vs-semrush/
Markdown: /en/comparison/verity-vs-semrush.md
FR HTML: /fr/comparatif/verity-vs-semrush/
FR Markdown: /fr/comparatif/verity-vs-semrush.md

Verity Score vs Profound: Shopify app with prompt tracking, fixes with undo and AI traffic to orders, compared with an enterprise platform. Start a free trial.
HTML: /en/comparison/verity-vs-profound/
Markdown: /en/comparison/verity-vs-profound.md
FR HTML: /fr/comparatif/verity-vs-profound/
FR Markdown: /fr/comparatif/verity-vs-profound.md

Verity Score vs Ahrefs Brand Radar: prompt tracking, fixes written to Shopify and AI traffic to orders for merchants. Compare both tools and start a free trial.
HTML: /en/comparison/verity-vs-ahrefs/
Markdown: /en/comparison/verity-vs-ahrefs.md
FR HTML: /fr/comparatif/verity-vs-ahrefs/
FR Markdown: /fr/comparatif/verity-vs-ahrefs.md

Verity Score vs Screaming Frog: AI prompt tracking, fixes written to Shopify and AI traffic to orders, compared with a desktop crawler. Start a free trial now.
HTML: /en/comparison/verity-vs-screaming-frog/
Markdown: /en/comparison/verity-vs-screaming-frog.md
FR HTML: /fr/comparatif/verity-vs-screaming-frog/
FR Markdown: /fr/comparatif/verity-vs-screaming-frog.md

Verity Score vs Otterly: prompt tracking, fixes written to Shopify and AI traffic to orders, versus Otterly AI search monitoring. Start your free trial.
HTML: /en/comparison/verity-vs-otterly/
Markdown: /en/comparison/verity-vs-otterly.md
FR HTML: /fr/comparatif/verity-vs-otterly/
FR Markdown: /fr/comparatif/verity-vs-otterly.md

Verity Score vs AthenaHQ: prompt tracking, Shopify fixes with undo and AI traffic to orders, versus a multi-CMS GEO platform. Start your free trial today.
HTML: /en/comparison/verity-vs-athenahq/
Markdown: /en/comparison/verity-vs-athenahq.md
FR HTML: /fr/comparatif/verity-vs-athenahq/
FR Markdown: /fr/comparatif/verity-vs-athenahq.md

Verity Score vs Peec AI: prompt tracking, fixes written to Shopify and AI traffic to orders, versus AI visibility analytics. Start your free trial on Shopify.
HTML: /en/comparison/verity-vs-peec-ai/
Markdown: /en/comparison/verity-vs-peec-ai.md
FR HTML: /fr/comparatif/verity-vs-peec-ai/
FR Markdown: /fr/comparatif/verity-vs-peec-ai.md

Verity Score vs Scrunch: prompt tracking, fixes written to Shopify and AI traffic to orders, versus Scrunch AI visibility monitoring. Start your free trial.
HTML: /en/comparison/verity-vs-scrunch/
Markdown: /en/comparison/verity-vs-scrunch.md
FR HTML: /fr/comparatif/verity-vs-scrunch/
FR Markdown: /fr/comparatif/verity-vs-scrunch.md

---


## GEO Glossary

20 essential terms for Generative Engine Optimization and AI commerce.

**ACP (Agentic Commerce Protocol)**  - Protocol developed by OpenAI and Stripe enabling AI agents to make purchases directly within ChatGPT. Standardizes product discovery, price comparison, and automated checkout through tool calls.

**AggregateRating**  - A schema.org type that exposes the average review score and review count directly in HTML code. Critical for GEO because AI crawlers cannot execute JavaScript from third-party review widgets.

**AI Buyer Score**  - A score simulating how an autonomous AI shopping agent would evaluate a product page. Assesses whether an agent can understand the product, compare prices, verify reviews, check shipping policies, and complete a purchase.

**AI Crawlers**  - Web crawling bots used by AI systems to index website content. Major AI crawlers include GPTBot (OpenAI), PerplexityBot, ClaudeBot (Anthropic), Google-Extended (Gemini), Amazonbot, and Applebot-Extended.

**AI Visibility**  - The ability of an e-commerce site to be discovered, understood, and recommended by AI systems. Depends on technical factors (robots.txt, schema.org, agent-card.json, sitemap), content quality, and trust signals.

**Agentic Commerce**  - An e-commerce model where autonomous AI agents search, compare products, and complete purchases on behalf of consumers using standardized protocols (ACP, UCP).

**ChatGPT Shopping**  - A ChatGPT feature enabling users to search, compare, and purchase products directly within the conversational interface. Merchants optimized for GEO appear in product recommendations.

**Claims and Proofs**  - In GEO, claims are marketing assertions (free shipping, money-back guarantee). Proofs are verifiable elements supporting them (accessible policy pages, certifications, schema.org data). AI systems only cite claims backed by proof.

**Conversational Content**  - A content strategy structuring product information as answers to natural questions. Generative AI systems favor this format because it matches their question-answer interaction pattern.

**GEO (Generative Engine Optimization)**  - The practice of optimizing an e-commerce store to be cited, compared, and recommended by generative search engines (ChatGPT, Perplexity, Google AI Mode) and AI shopping agents.

**GEO Score**  - A score out of 100 measuring a store's ability to be cited and recommended by generative AI. Combines 9 factors: AI crawler accessibility, schema.org completeness, AggregateRating, native HTML content, verifiable proofs, claims coherence, agent discovery, sitemap health, and canonical consistency.

**Google AI Mode**  - Google's search mode generating Gemini-powered answers instead of listing links. Uses the Shopping Graph and UCP protocol to recommend products with integrated price comparison.

**JSON-LD**  - Google's recommended format for structured data serialization. In e-commerce, JSON-LD is injected in a script tag containing Product, Offer, and AggregateRating data readable by crawlers without parsing the DOM.

**Perplexity Shopping**  - Perplexity's e-commerce feature displaying products with prices, images, and purchase links in answers. Uses structured data and server-rendered content to populate product cards.

**Recommendability**  - The likelihood that a store will be recommended by a generative AI in response to a shopping query. Combines complete structured data, verifiable reviews, competitive pricing, clear policies, and frictionless checkout.

**Schema.org**  - A standardized structured data vocabulary. For e-commerce, Product, Offer, AggregateRating, and Brand types enable AI engines to instantly understand product attributes without parsing raw HTML.

**Shopping Graph**  - Google's database of billions of structured product listings. Powers Google Shopping, Google AI Mode, and Gemini product responses. Fed through Merchant Center, schema.org markup, and UCP.

**SSR (Server-Side Rendering)**  - A rendering technique where HTML is generated on the server before being sent to the browser. Critical for GEO because AI crawlers can only read initial HTML.

**UCP (Universal Commerce Protocol)**  - Protocol co-developed by Google, Shopify, and other platforms to enable AI agents to discover and purchase products. Exposes a .well-known/ucp manifest describing transactional capabilities.

**llms.txt**  - A text file placed at a website's root that acts as a compact AI content index. It can orient agents toward collections, policies and proof pages, but remains secondary to canonical HTML, schema.org, sitemap coverage and agent-card.json.

---


---

# Muse de Meta et Shopify : ce qui est vraiment disponible
> Muse, Shopify et Shop Pay : catalogue, paiement, éligibilité et attribution. Analyse des sources officielles Meta et Shopify consultées le 9 septembre 2026.
- Canonical HTML: https://verityscore.io/fr/blog/muse-meta-shopify/
- Markdown alternate: https://verityscore.io/fr/blog/muse-meta-shopify.md
- Language: fr
- Content type: blog
- Published: 2026-09-09
- Updated: 2026-09-09
- Tags: shopify, muse, meta, ai-commerce, product-data
<div class="muse-review-content">


<p>Pour un marchand Shopify, la priorité est de distinguer quatre états : l&#39;annonce d&#39;un canal, l&#39;accès aux données du catalogue, la possibilité de payer et l&#39;enregistrement d&#39;une commande. Pour une équipe SEO ou acquisition, un cinquième état compte : pouvoir relier cette commande au parcours qui l&#39;a produite. Une présence dans un catalogue n&#39;apporte aucune de ces preuves à elle seule.</p>


<section class="muse-ecosystem" aria-labelledby="muse-ecosystem-title">
<h2 id="muse-ecosystem-title">Les entreprises derrière l’écosystème Muse</h2>
<div class="muse-logo-grid">
<div class="muse-company"><div class="muse-company-logo"><img src="/images/articles/muse-meta-shopify/meta-logo.png" width="116" height="50" alt="Meta"><img class="ecosystem-muse" src="/images/articles/muse-meta-shopify/muse-logo.svg" width="32" height="32" alt="Muse"></div><span>Muse</span></div>
<div class="muse-company"><div class="muse-company-logo"><img src="/images/articles/muse-meta-shopify/stripe-logo.svg" width="96" height="40" alt="Stripe"></div><span>Link</span></div>
<div class="muse-company"><div class="muse-company-logo"><img class="ecosystem-icon" src="/shopify-bag.svg" width="36" height="42" alt=""><strong>Shopify</strong></div><span>Shop Pay · annoncé à venir</span></div>
<div class="muse-company"><div class="muse-company-logo"><img class="ecosystem-icon" src="/images/articles/muse-meta-shopify/1password-logo.png" width="42" height="42" alt=""><strong>1Password</strong></div><span>Prise en charge annoncée</span></div>
</div>
<p class="muse-logo-source">Entreprises et services cités dans l’annonce officielle de Meta du 8 septembre 2026. <a href="#ref-1">Source [1]</a></p>
</section>

## Ce que Muse change dans le parcours d'achat


<p>Meta décrit Muse comme un agent personnel capable de travailler dans la durée, d&#39;utiliser un navigateur et de poursuivre une tâche après la fermeture de l&#39;application. Son intérêt commercial dépasse donc la réponse à une recherche de produit. L&#39;agent peut préparer une action, revenir lorsqu&#39;une condition change et demander une validation avant l&#39;achat. L&#39;annonce prévoit l&#39;application Muse et WhatsApp comme interfaces de conversation. Elle situe le déploiement initial aux États-Unis, sur iOS, Android et le Web. <a href="#ref-1" aria-label="Source 1">[1]</a></p>
<p>La documentation technique précise le parcours de paiement décrit au lancement : Link fournit une carte à usage unique, liée à un marchand, à un montant et à une durée de validité. L&#39;agent demande une approbation humaine. Shop Pay y apparaît également comme une intégration à venir. Il s&#39;agit de capacités déclarées par Meta, sans test d&#39;achat réalisé pour cette analyse. <a href="#ref-2" aria-label="Source 2">[2]</a></p>
<figure class="media-card">
<div class="media-label muse-brand-label"><img src="/images/articles/muse-meta-shopify/muse-logo.svg" width="36" height="36" alt="Logo officiel Muse de Meta"> Démonstration officielle · Muse de Meta</div>
<video poster="/images/articles/muse-meta-shopify/muse-video-poster.png" controls playsinline preload="none" aria-label="Démonstration shopping de Muse publiée par Meta" width="960" height="836"><source src="https://about.fb.com/wp-content/uploads/2026/09/Muse_Shopping.mp4" type="video/mp4"></video>
<figcaption>Meta illustre un scénario de shopping dans son annonce du 8 septembre 2026. Cette démonstration de l’éditeur ne constitue pas un achat testé pour cet article. Le texte ci-dessus décrit le parcours documenté. <a href="https://about.fb.com/news/2026/09/introducing-muse-personal-ai-agent/">Source et contexte</a> · <a href="https://about.fb.com/wp-content/uploads/2026/09/Muse_Shopping.mp4">Ouvrir la vidéo</a>.</figcaption>
</figure>

<aside class="social-card" aria-label="Publication de lancement Muse">
<div class="media-label">À la source · publication sur X</div>
<p>Le compte Muse annonce son lancement le 8 septembre 2026. Ce post établit la communication de lancement, sans prouver l’accès ou le paiement pour une boutique donnée.</p>
<iframe class="muse-tweet" src="https://platform.twitter.com/embed/Tweet.html?id=2097399178376671666&amp;dnt=true&amp;lang=fr&amp;theme=light&amp;width=550px" title="Publication officielle de Muse sur X" referrerpolicy="no-referrer" loading="eager" width="550" height="780"></iframe><a class="source-link" href="https://x.com/Muse/status/2097399178376671666" rel="nofollow noopener noreferrer" target="_blank">Consulter la publication sur X ↗</a></aside>

<p>Stripe apporte un recoupement utile dans son communiqué du 8 septembre : les consommateurs américains peuvent connecter Link à Muse. Il décrit un paiement avec le moyen enregistré chez les marchands acceptant Link, et une carte virtuelle à usage unique pour les autres achats pris en charge. Le montant doit être approuvé dans la conversation. Cette confirmation porte sur Link, sans établir la disponibilité de Shop Pay dans Muse. <a href="#ref-17" aria-label="Source 17">[17]</a></p>
<p>L&#39;implication pour une boutique est concrète : un agent qui navigue peut rencontrer la fiche produit, les conditions de livraison et le paiement du site. Un canal qui propose un paiement intégré peut présenter une autre interface. Les données, les personnalisations et la mesure ne suivent pas nécessairement le même chemin. Il faut identifier le parcours avant de conclure qu&#39;un changement de catalogue, de thème ou de tracking répondra au problème.</p>
<div class="scroll" tabindex="0" role="region" aria-label="Tableau, défilement horizontal possible"><table class="states">
<thead>
<tr>
<th>État</th>
<th>Ce que l&#39;on peut établir</th>
<th>Ce qui reste à prouver</th>
</tr>
</thead>
<tbody><tr>
<td>Annonce</td>
<td>Une plateforme décrit une capacité et un calendrier</td>
<td>L&#39;accès d&#39;un utilisateur donné</td>
</tr>
<tr>
<td>Catalogue</td>
<td>Des données sont accessibles à un canal</td>
<td>Leur sélection et leur affichage pour une requête</td>
</tr>
<tr>
<td>Paiement</td>
<td>Un parcours est documenté pour un périmètre éligible</td>
<td>Son fonctionnement pour le produit, le pays et la boutique</td>
</tr>
<tr>
<td>Commande</td>
<td>Une commande figure dans le système marchand</td>
<td>Son origine précise, son paiement et son résultat après retours</td>
</tr>
</tbody></table></div>
<figure class="muse-diagram"><div class="diagram-heading"><span class="diagram-index">01</span><h3>Quatre états, quatre preuves</h3></div><div class="diagram-grid"><div class="diagram-card"><span class="diagram-step">01</span><h4>Annonce</h4><p>Un canal est présenté par la plateforme.</p></div><div class="diagram-card"><span class="diagram-step">02</span><h4>Catalogue</h4><p>Les données produit sont accessibles.</p></div><div class="diagram-card"><span class="diagram-step">03</span><h4>Paiement</h4><p>Un parcours de paiement est disponible.</p></div><div class="diagram-card"><span class="diagram-step">04</span><h4>Commande</h4><p>Un achat est effectivement enregistré.</p></div></div><figcaption>Une annonce ne prouve ni l’accès marchand, ni le paiement, ni la commande. Schéma de lecture, sans données de performance.<a class="figure-brand" href="/fr/blog/"><img src="/logo.svg" width="20" height="20" alt="">Articles de Verity Score</a></figcaption></figure>
<p>Cette grille est une méthode de lecture, pas une mesure de performance de Muse. Elle évite notamment de transformer « produits disponibles » en « ventes réalisées ».</p>


## Shop Pay : une divergence qui impose de nommer la surface


<p>La page Meta du 8 septembre indique explicitement que Shop Pay arrive prochainement. La publication technique sur la sécurité répète cette réserve. En parallèle, la page d&#39;aide Shopify « Selling on Meta » décrit un paiement direct activé par défaut pour les boutiques éligibles. Elle parle de surfaces Meta, sans établir un calendrier propre à Muse et Shop Pay. <a href="#ref-1" aria-label="Source 1">[1]</a><a href="#ref-2" aria-label="Source 2">[2]</a><a href="#ref-3" aria-label="Source 3">[3]</a></p>
<figure class="source-capture">
<a href="https://about.fb.com/news/2026/09/introducing-muse-personal-ai-agent/"><img src="/images/articles/muse-meta-shopify/capture-meta-shop-pay.png" width="215" height="27" alt="Extrait exact de Meta : Shop Pay is coming soon"></a>
<figcaption>Capture réelle du passage de l’annonce Meta, recadrée sur la mention Shop Pay, consultée le 9 septembre 2026. Traduction : Shop Pay arrive prochainement. <a href="https://about.fb.com/news/2026/09/introducing-muse-personal-ai-agent/">Relire le paragraphe complet chez Meta</a>.</figcaption>
</figure>

<figure class="muse-diagram"><div class="diagram-heading"><span class="diagram-index">02</span><h3>Shop Pay : deux périmètres à distinguer</h3></div><div class="diagram-grid"><div class="diagram-card"><span class="diagram-step">01</span><h4><img class="brand-muse" src="/images/articles/muse-meta-shopify/muse-logo.svg" width="28" height="28" alt="Logo Muse de Meta"> Meta · Muse</h4><p>L’annonce Meta indique que Shop Pay arrive prochainement.</p></div><div class="diagram-card"><span class="diagram-step">02</span><h4><img class="brand-shopify" src="/shopify-bag.svg" width="24" height="28" alt=""> Shopify · surfaces Meta</h4><p>Le centre d’aide documente un paiement direct pour les boutiques éligibles.</p></div></div><figcaption>Les deux sources ne décrivent pas le même périmètre. Aucun calendrier général de Shop Pay dans Muse ne peut en être déduit.<a class="figure-brand" href="/fr/blog/"><img src="/logo.svg" width="20" height="20" alt="">Articles de Verity Score</a></figcaption></figure>
<p>Une différence de calendrier, de parcours ou de population pourrait expliquer l&#39;écart. Ce sont des hypothèses. Les sources consultées ne permettent pas de choisir entre elles. La formulation défendable est donc la suivante : <strong>Shopify documente le paiement direct Meta ; la disponibilité générale de Shop Pay dans Muse n&#39;est pas confirmée par l&#39;annonce Meta consultée.</strong></p>
<div class="scroll" tabindex="0" role="region" aria-label="Tableau, défilement horizontal possible"><table>
<thead>
<tr>
<th>Repère daté</th>
<th>Source</th>
<th>Portée réelle</th>
</tr>
</thead>
<tbody><tr>
<td>11 janvier 2026</td>
<td>Shopify annonce UCP, de nouvelles intégrations et le forfait Agentic</td>
<td>Contexte d&#39;infrastructure ; aucune preuve du lancement Muse</td>
</tr>
<tr>
<td>11 mai 2026</td>
<td>Changelog Shopify : nouvelle page Agentic dans l&#39;administration</td>
<td>Centralisation des réglages et de la mesure</td>
</tr>
<tr>
<td>7 juillet 2026</td>
<td>Date affichée des conditions Agentic Storefronts relues</td>
<td>Cadre contractuel ; pas date de disponibilité de chaque canal</td>
</tr>
<tr>
<td>8 septembre 2026</td>
<td>Meta présente Muse</td>
<td>Déploiement US, Link décrit, Shop Pay à venir</td>
</tr>
<tr>
<td>9 septembre 2026</td>
<td>Consultation des documents Shopify et Meta</td>
<td>Photographie documentaire, sans commande test</td>
</tr>
</tbody></table></div>
<p>Sources de la chronologie : Shopify, annonces et conditions ; Meta, annonce de lancement. Une documentation sans date de mise à jour visible n&#39;est pas datée artificiellement du jour de sa consultation. <a href="#ref-1" aria-label="Source 1">[1]</a><a href="#ref-8" aria-label="Source 8">[8]</a><a href="#ref-9" aria-label="Source 9">[9]</a><a href="#ref-10" aria-label="Source 10">[10]</a></p>


## Qui est éligible, et à quoi ?


<p>Shopify indique que le paiement direct Meta est affiché aux clients situés aux États-Unis. La condition formulée est de vendre à des clients américains. Cela ne permet ni d&#39;annoncer une disponibilité pour les acheteurs français, ni de conclure que tous les marchands établis en France sont exclus. Le pays de l&#39;acheteur et le pays d&#39;établissement du vendeur sont deux informations différentes. <a href="#ref-3" aria-label="Source 3">[3]</a></p>
<p>L&#39;éligibilité au catalogue constitue un autre filtre. Shopify exige notamment un forfait Starter ou supérieur, avec une exception documentée pour le forfait Agentic, une boutique hors mode privé, un titre, au moins une image et un prix supérieur à zéro. La publication sur les canaux compatibles et une URL produit identifiable comptent aussi. Les produits non répertoriés ou masqués des moteurs sont exclus. Shopify examine également la situation du compte et son historique ; une fiche correctement remplie ne garantit donc pas l&#39;admission de la boutique. <a href="#ref-4" aria-label="Source 4">[4]</a></p>
<div class="scroll" tabindex="0" role="region" aria-label="Tableau, défilement horizontal possible"><table>
<thead>
<tr>
<th>Situation</th>
<th>État documenté au 9 septembre</th>
<th>Conséquence pour le marchand</th>
</tr>
</thead>
<tbody><tr>
<td>Acheteur aux États-Unis</td>
<td>Déploiement Muse annoncé ; paiement direct Meta prévu pour clients US</td>
<td>Vérifier l&#39;accès et le parcours précis</td>
</tr>
<tr>
<td>Acheteur en France</td>
<td>L&#39;annonce consultée ne confirme pas de lancement français</td>
<td>Ne pas présenter Muse comme canal local disponible</td>
</tr>
<tr>
<td>Boutique éligible avec produits éligibles</td>
<td>Paiement direct Meta décrit comme actif par défaut</td>
<td>Lire l&#39;état réel du canal dans l&#39;administration</td>
</tr>
<tr>
<td>Abonnement, lot, produit personnalisable ou exclusivement B2B</td>
<td>Non pris en charge dans le paiement direct Meta</td>
<td>Ne pas étendre l&#39;éligibilité catalogue à ce paiement</td>
</tr>
<tr>
<td>Forfait Agentic sans boutique Shopify principale</td>
<td>Procédure spécifique ; paiement direct désactivé par défaut</td>
<td>Ne pas appliquer les automatismes des autres forfaits</td>
</tr>
<tr>
<td>Publicité achetable Meta</td>
<td>Canal Facebook et Instagram avec synchronisation du flux requis</td>
<td>Distinguer ce flux de la découverte via Catalog</td>
</tr>
</tbody></table></div>
<p>Les restrictions de paiement proviennent de la page Meta de Shopify, les règles de catalogue de sa page d&#39;éligibilité, et l&#39;exception du forfait Agentic de son guide dédié. Ce tableau n&#39;est pas un verdict sur une boutique particulière. <a href="#ref-3" aria-label="Source 3">[3]</a><a href="#ref-4" aria-label="Source 4">[4]</a><a href="#ref-7" aria-label="Source 7">[7]</a></p>
<p>Les limites touchent aussi l&#39;expérience. Shopify exclut du paiement direct Meta certains blocs de checkout ainsi que la livraison locale, le retrait en magasin et les points de retrait. Les remises automatiques et codes de réduction sont décrits comme pris en charge. Les Shopify Functions relatives, entre autres, à la livraison et à la validation peuvent fonctionner, mais cela ne prouve pas que chaque personnalisation du site se retrouve dans Meta. <a href="#ref-3" aria-label="Source 3">[3]</a></p>


## Catalogue : l'intégration réduit un travail, elle n'efface pas les contrôles


<p>La page Shopify consacrée à Meta affirme que la découverte s&#39;appuie sur Shopify Catalog. Elle rattache explicitement l&#39;obligation du flux Facebook et Instagram aux publicités achetables. Une autre page, consacrée à la découverte produit, décrit plus largement les surfaces Meta comme alimentées par ce canal Facebook et Instagram. La documentation n&#39;est donc pas parfaitement alignée dans sa formulation. <a href="#ref-3" aria-label="Source 3">[3]</a><a href="#ref-5" aria-label="Source 5">[5]</a></p>
<p>Pour traiter cette divergence, la page spécifique Meta est la référence la plus précise sur la distinction découverte/publicité. Elle ne justifie pas d&#39;installer un canal publicitaire uniquement parce qu&#39;un email mentionne Muse. Pour une boutique donnée, le bon contrôle consiste à examiner la surface voulue, son état dans Agentic et ses conditions actuelles. Une intégration existante n&#39;est pas à modifier à partir d&#39;une généralisation tirée d&#39;une autre page.</p>
<p>Shopify décrit Catalog comme une source structurée de titres, descriptions, options, images, prix et disponibilité, mise à jour en continu. Cette promesse ne fournit pas un délai maximal mesuré entre une modification de stock et sa restitution dans Muse. Elle ne démontre pas davantage que la bonne variante sera toujours retenue pour une demande ambiguë. <a href="#ref-5" aria-label="Source 5">[5]</a></p>
<p>Une équipe ecommerce peut préparer une vérification sans rien changer : choisir une référence et une variante, noter leur prix et leur devise, relever la disponibilité pour le marché ciblé, puis comparer les informations à chaque étape accessible du parcours. Si une observation Muse devient possible, elle doit porter le même horodatage et le même contexte. Un écart est à conserver comme écart observé ; sa cause n&#39;est pas automatiquement une mauvaise synchronisation Shopify.</p>
<p>Pour les catalogues qui utilisent des champs personnalisés, le sujet peut être le mapping : la donnée existe dans la boutique, mais n&#39;est pas reprise dans le champ attendu par Catalog. Shopify documente cette possibilité. En SEO, le travail utile consiste alors à rendre les attributs réellement disponibles et compréhensibles, avec les contraintes d&#39;usage, les dimensions ou les compatibilités qui permettent une décision. Ajouter des adjectifs ou rallonger une description sans information n&#39;apporte pas cette preuve. <a href="#ref-5" aria-label="Source 5">[5]</a></p>
<p>La méthode Verity sur les <a href="https://verityscore.io/fr/blog/chatgpt-shopify-ecart-representation-produit/">écarts de représentation produit</a> fournit un cadre complémentaire pour comparer fait approuvé et affichage. Son terrain est ChatGPT ; ses constats propres à cette plateforme ne sont pas des résultats mesurés sur Muse. <a href="#ref-13" aria-label="Source 13">[13]</a></p>
<aside class="muse-reading"><span class="media-label">Pour préparer votre catalogue</span><p>Pour prolonger ces vérifications, le <a href="https://universalcommerceprotocol.blog/catalogues-produits-pour-ia/">guide des catalogues produits pour les agents IA</a> propose une grille de lecture des identifiants, attributs, prix, stocks et politiques de livraison ou de retour. C’est un complément méthodologique ; il ne démontre pas une intégration de Muse au protocole UCP.</p></aside>


## Une commande Meta ne mesure pas à elle seule l'apport de Muse


<p>Shopify indique que les commandes passées sur les surfaces Meta sont attribuées à Meta dans l&#39;administration. La documentation consultée ne démontre pas l&#39;existence d&#39;un champ qui isole Muse dans toutes ces commandes. Une hausse du canal Meta ne peut donc pas devenir automatiquement une hausse des ventes Muse. <a href="#ref-3" aria-label="Source 3">[3]</a></p>
<p>Le paiement direct introduit une autre rupture de mesure. Shopify précise que Google Analytics et les pixels personnalisés ne se déclenchent pas dans ce checkout. Les événements décrits sont transmis de serveur à serveur pour le début et la fin du paiement. Le responsable acquisition doit éviter de comparer sans précaution ce parcours à une conversion web mesurée par des pixels côté navigateur. <a href="#ref-3" aria-label="Source 3">[3]</a></p>
<p>La page Agentic présente des rapports de ventes, commandes, sessions de la boutique et conversions de ces sessions. Ces mesures ne partagent pas toutes le même périmètre. Diviser toutes les commandes attribuées à un canal par les seules sessions de la boutique peut produire un taux sans sens si certaines commandes ont été finalisées hors du site. <a href="#ref-6" aria-label="Source 6">[6]</a></p>
<figure class="muse-diagram"><div class="diagram-heading"><span class="diagram-index">03</span><h3>De la visibilité à la commande</h3></div><div class="diagram-grid"><div class="diagram-card"><span class="diagram-step">01</span><h4>Visibilité</h4><p>Le produit apparaît dans une réponse ou un catalogue.</p></div><div class="diagram-card"><span class="diagram-step">02</span><h4>Visite</h4><p>Une session est mesurée sur la boutique.</p></div><div class="diagram-card"><span class="diagram-step">03</span><h4>Commande attribuée</h4><p>Une vente est reliée à une source selon les données disponibles.</p></div></div><figcaption>Ces mesures répondent à des questions différentes. Aucun chiffre de trafic, de conversion ou de revenu Muse n’est représenté.<a class="figure-brand" href="/fr/blog/"><img src="/logo.svg" width="20" height="20" alt="">Articles de Verity Score</a></figcaption></figure>
<p>Il faut conserver, pour chaque analyse, la période, le canal ou référent disponible, le lieu de finalisation connu et les commandes annulées ou remboursées. Si l&#39;origine Muse n&#39;est pas exposée, la catégorie reste « Meta, surface non isolée ». Cette discipline laisse une inconnue visible plutôt que de fabriquer un résultat.</p>
<p>Le <a href="https://verityscore.io/fr/blog/share-of-ai-voice-guide-2026/">guide Verity sur le Share of AI Voice</a> est utile pour définir la visibilité comme un indicateur distinct. Une présence dans une réponse, une visite et une commande ont des dénominateurs différents. Aucun volume de trafic, taux de conversion ou revenu Muse n&#39;a été mesuré dans cette analyse. <a href="#ref-14" aria-label="Source 14">[14]</a></p>


## Frais, retours et données : ce que conserve le marchand


<p>La page Meta de Shopify indique qu&#39;aucuns frais spécifiques ne s&#39;ajoutent pour vendre dans son paiement direct, en dehors des frais habituels de traitement du paiement. « Sans frais supplémentaires » ne signifie donc pas « transaction gratuite ». Les conditions Agentic Storefronts prévoient par ailleurs la possibilité de frais selon les canaux, présentés dans l&#39;administration. La règle actuelle documentée pour Meta ne constitue pas une garantie de gratuité permanente pour tous les canaux. <a href="#ref-3" aria-label="Source 3">[3]</a><a href="#ref-8" aria-label="Source 8">[8]</a></p>
<p>Le marchand reste le vendeur de référence, ou merchant of record. Il gère les commandes, leur exécution, les retours et le service client. Un achat engagé dans une conversation ne transfère pas automatiquement ces responsabilités à l&#39;agent. Les politiques de la boutique doivent être présentes et à jour ; elles font partie des conditions publiées par Shopify. <a href="#ref-3" aria-label="Source 3">[3]</a></p>
<p>Meta mentionne aussi les protections Link. La page Link consultée les réserve aux achats éligibles et précise que Cover Genius évalue la couverture. Il serait erroné de les promettre pour toute commande Shopify ou pour un futur paiement Shop Pay. Le service client doit d&#39;abord identifier le moyen de paiement, le parcours et les conditions applicables au dossier. <a href="#ref-1" aria-label="Source 1">[1]</a><a href="#ref-11" aria-label="Source 11">[11]</a></p>
<p>Côté données, Shopify décrit l&#39;accès des canaux aux informations produit et, lorsqu&#39;une commande est passée dans un paiement direct, aux données nécessaires telles que le nom, l&#39;email ou l&#39;adresse du client. La documentation indique que l&#39;ensemble de l&#39;historique des commandes et de la base clients n&#39;est pas ouvert au canal. Cela reste différent d&#39;une absence de partage : les données nécessaires à la transaction circulent. <a href="#ref-12" aria-label="Source 12">[12]</a></p>


## Les limites de confidentialité de Muse sont déjà documentées


<p>Meta décrit une machine virtuelle dédiée à chaque utilisateur, un stockage isolé des identifiants et un agent Sentinel qui contrôle les actions vers l&#39;extérieur. La même publication reconnaît que Muse peut commettre des erreurs et subir des attaques par injection d&#39;instructions. Ces descriptions constituent l&#39;architecture déclarée par son éditeur ; elles ne valent pas certification indépendante dans cette analyse. <a href="#ref-2" aria-label="Source 2">[2]</a></p>
<p>La distinction entre les versions de la machine virtuelle est essentielle. Meta indique que l&#39;architecture actuelle peut permettre un accès à certaines données pour exploiter, sécuriser ou assister le service. La version Confidential VM, censée empêcher cryptographiquement cet accès, est annoncée pour plus tard dans l&#39;année. Présenter cette propriété future comme acquise au lancement serait inexact. <a href="#ref-2" aria-label="Source 2">[2]</a></p>
<p>La politique publicitaire demande la même précision. Meta dit ne pas transmettre les conversations ou les données de la VM à ses systèmes publicitaires. Mais sa documentation précise que les visites réalisées par Muse sur des sites peuvent être traitées comme l&#39;activité de l&#39;utilisateur et influencer les annonces qu&#39;il voit ensuite. L&#39;absence de partage direct des conversations ne signifie pas l&#39;absence de toute conséquence publicitaire du parcours. <a href="#ref-2" aria-label="Source 2">[2]</a></p>
<p>Enfin, Meta explique utiliser des trajectoires d&#39;interaction nettoyées de certaines informations personnelles pour entraîner ses modèles, avec une option de refus dans les réglages. Il faut distinguer ce mécanisme du stockage des identifiants, du partage publicitaire et des données de commande accessibles aux partenaires. Ce sont des traitements différents, soumis à des contrôles différents. <a href="#ref-2" aria-label="Source 2">[2]</a></p>


## Les cinq décisions à prendre avant d'augmenter l'effort acquisition


<div class="scroll" tabindex="0" role="region" aria-label="Tableau, défilement horizontal possible"><table>
<thead>
<tr>
<th>Priorité proposée</th>
<th>Responsable</th>
<th>Vérification utile</th>
<th>Décision permise par la preuve</th>
</tr>
</thead>
<tbody><tr>
<td>Qualifier le périmètre</td>
<td>Ecommerce</td>
<td>Pays acheteur, produit, canal et paiement activé</td>
<td>Identifier un parcours candidat, sans annoncer une disponibilité universelle</td>
</tr>
<tr>
<td>Contrôler une référence</td>
<td>Catalogue et SEO</td>
<td>Variante, prix, devise, stock, attributs et politiques</td>
<td>Corriger une donnée manquante ou contradictoire identifiée</td>
</tr>
<tr>
<td>Examiner les exceptions</td>
<td>Ecommerce et technique</td>
<td>Lots, personnalisation, abonnements, livraison, fonctions du checkout</td>
<td>Éviter une promesse que le parcours direct ne peut pas tenir</td>
</tr>
<tr>
<td>Définir la mesure</td>
<td>Acquisition et analytics</td>
<td>Attribution disponible, lieu de paiement, sessions, annulations et retours</td>
<td>Distinguer observation de canal et revenu incrémental</td>
</tr>
<tr>
<td>Relire les contrôles</td>
<td>Responsable de boutique</td>
<td>Catalog, direct checkout, inscription future et conditions</td>
<td>Choisir les canaux en connaissance de leurs effets</td>
</tr>
</tbody></table></div>
<p>Ce plan est une recommandation. Il commence par des lectures et des comparaisons réversibles. Il ne fixe pas un délai universel ni un budget publicitaire : ceux-ci dépendent du catalogue, de l&#39;accès au produit et des résultats réellement observés.</p>
<p>Les réglages se trouvent dans Canaux de vente &gt; Agentic. Shopify permet de gérer l&#39;accès au catalogue et le paiement direct séparément lorsque la gestion automatique est désactivée. Retirer Catalog peut prendre jusqu&#39;à sept jours pour que le partage cesse, selon sa documentation, et n&#39;empêche pas nécessairement la découverte par exploration du Web. Masquer entièrement un produit peut affecter sa visibilité dans les moteurs ; ce n&#39;est pas une simple préférence Muse. <a href="#ref-6" aria-label="Source 6">[6]</a></p>
<p>Le lancement justifie de préparer les données et la mesure. Il ne justifie pas encore de présenter Muse comme une source de revenus prouvée pour une boutique. Le prochain résultat utile est une trace vérifiable reliant un produit, un acheteur éligible, un parcours de paiement et une commande attribuable. Tant que l&#39;un de ces éléments manque, il doit rester nommé.</p>


## Sources et méthode


<p>Analyse documentaire arrêtée au 9 septembre 2026. Sources officielles recherchées puis relues, affirmations confrontées par surface et date. Aucune boutique, messagerie ou donnée client consultée ; aucun achat, réglage ou test terrain. Les hypothèses et recommandations sont distinguées des déclarations des plateformes. Les annexes locales conservent les extractions, contradictions et contrôles.</p>
<p id="ref-1"><a href="#ref-1" aria-label="Source 1">[1]</a> Meta. <a rel="nofollow noopener noreferrer" href="https://about.fb.com/news/2026/09/introducing-muse-personal-ai-agent/">Introducing Muse: The World’s First Personal AI Agent Built for Everyone</a>. 8 septembre 2026. Annonce de déploiement, pas preuve d&#39;accès universel.</p>
<p id="ref-2"><a href="#ref-2" aria-label="Source 2">[2]</a> Meta. <a rel="nofollow noopener noreferrer" href="https://security.muse.ai/">How We Built Safety Into Muse</a>. 8 septembre 2026 selon les métadonnées ; consulté le 9. Architecture, achats, données et limites déclarés par l&#39;éditeur.</p>
<p id="ref-3"><a href="#ref-3" aria-label="Source 3">[3]</a> Shopify Help Center. <a rel="nofollow noopener noreferrer" href="https://help.shopify.com/en/manual/online-sales-channels/agentic-storefronts/meta">Selling on Meta</a>. Consulté le 9 septembre 2026 ; date de mise à jour non affichée dans le corps consulté. Portée Meta, sans calendrier Muse spécifique.</p>
<p id="ref-4"><a href="#ref-4" aria-label="Source 4">[4]</a> Shopify Help Center. <a rel="nofollow noopener noreferrer" href="https://help.shopify.com/en/manual/shopify-catalog/requirements">Requirements for being included in Shopify Catalog</a>. Consulté le 9 septembre 2026. Conditions de catalogue, pas garantie de recommandation.</p>
<p id="ref-5"><a href="#ref-5" aria-label="Source 5">[5]</a> Shopify Help Center. <a rel="nofollow noopener noreferrer" href="https://help.shopify.com/en/manual/online-sales-channels/agentic-storefronts/products">Shopify Catalog and product discovery for agentic storefronts</a>. Consulté le 9 septembre 2026. Divergence de formulation sur les surfaces Meta conservée.</p>
<p id="ref-6"><a href="#ref-6" aria-label="Source 6">[6]</a> Shopify Help Center. <a rel="nofollow noopener noreferrer" href="https://help.shopify.com/en/manual/online-sales-channels/agentic-storefronts/agentic-home">Managing agentic storefronts</a>. Consulté le 9 septembre 2026. Réglages et rapports décrits, sans vérification d&#39;un compte.</p>
<p id="ref-7"><a href="#ref-7" aria-label="Source 7">[7]</a> Shopify Help Center. <a rel="nofollow noopener noreferrer" href="https://help.shopify.com/en/manual/online-sales-channels/agentic-storefronts/agentic-plan-setup">Setting up agentic storefronts for stores on the Agentic plan</a>. Consulté le 9 septembre 2026. Procédure d&#39;un forfait distinct.</p>
<p id="ref-8"><a href="#ref-8" aria-label="Source 8">[8]</a> Shopify. <a rel="nofollow noopener noreferrer" href="https://www.shopify.com/legal/terms-agentic-storefronts">Agentic Storefronts Supplemental Terms of Service</a>. Mise à jour affichée : 7 juillet 2026, lecture fraîche le 9 septembre. Une extraction antérieure renvoyant le 25 mai a été écartée pour la date et la version.</p>
<p id="ref-9"><a href="#ref-9" aria-label="Source 9">[9]</a> Shopify. <a rel="nofollow noopener noreferrer" href="https://www.shopify.com/news/ai-commerce-at-scale">The agentic commerce platform</a>. 11 janvier 2026. Contexte historique, sans projection des capacités UCP sur Muse.</p>
<p id="ref-10"><a href="#ref-10" aria-label="Source 10">[10]</a> Shopify Changelog. <a rel="nofollow noopener noreferrer" href="https://changelog.shopify.com/posts/a-new-home-for-agentic-storefronts-in-your-admin">A new home for Agentic Storefronts in your admin</a>. 11 mai 2026.</p>
<p id="ref-11"><a href="#ref-11" aria-label="Source 11">[11]</a> Link. <a rel="nofollow noopener noreferrer" href="https://support.link.com/questions/what-s-covered-with-protections">What’s covered with protections?</a>. Consulté le 9 septembre 2026. Couverture conditionnelle ; aucun contrat individuel vérifié.</p>
<p id="ref-12"><a href="#ref-12" aria-label="Source 12">[12]</a> Shopify Help Center. <a rel="nofollow noopener noreferrer" href="https://help.shopify.com/en/manual/online-sales-channels/agentic-storefronts/data-privacy">Data sharing and privacy for selling with agentic storefronts</a>. Consulté le 9 septembre 2026.</p>
<p id="ref-13"><a href="#ref-13" aria-label="Source 13">[13]</a> Verity Score. <a rel="nofollow noopener noreferrer" href="https://verityscore.io/fr/blog/chatgpt-shopify-ecart-representation-produit/">ChatGPT et Shopify : auditer la représentation produit</a>. 31 août 2026 ; relu le 9 septembre. Complément méthodologique, pas preuve Muse.</p>
<p id="ref-14"><a href="#ref-14" aria-label="Source 14">[14]</a> Verity Score. <a rel="nofollow noopener noreferrer" href="https://verityscore.io/fr/blog/share-of-ai-voice-guide-2026/">Share of AI Voice : guide mesure 2026</a>. Consulté le 9 septembre 2026. Complément de définition, sans réutilisation de ses métriques.</p>
<p id="ref-15"><a href="#ref-15" aria-label="Source 15">[15]</a> Meta. <a rel="nofollow noopener noreferrer" href="https://introducing.muse.ai/">How We Designed Muse</a>. Septembre 2026. Référence de conception lue pour le contexte ; anecdotes de l&#39;équipe non reprises comme observations indépendantes.</p>
<p id="ref-16"><a href="#ref-16" aria-label="Source 16">[16]</a> Shopify. <a rel="nofollow noopener noreferrer" href="https://www.shopify.com/agentic-storefronts">Sell in AI chats with Shopify</a>. Consulté le 9 septembre 2026. Page commerciale confrontée aux conditions plus précises du centre d&#39;aide.</p>

<p id="ref-17"><a href="#ref-17" aria-label="Source 17">[17]</a> Stripe. <a rel="nofollow noopener noreferrer" href="https://stripe.com/en-ca/newsroom/news/stripe-helps-meta-muse-shop-with-link">Stripe helps Muse shop across the internet with Link</a>. 8 septembre 2026 ; consulté le 9 septembre. Communiqué de la plateforme de paiement, sans achat indépendant.</p>

<p id="ref-18">[18] Universal Commerce Protocol Blog. <a href="https://universalcommerceprotocol.blog/catalogues-produits-pour-ia/">Guide des catalogues produits pour les agents IA</a>. Consulté le 9 septembre 2026. Complément méthodologique sur la structure du catalogue ; ne prouve pas une intégration UCP dans Muse.</p>


</div>


# Meta's Muse and Shopify: What Is Really Available
> Muse, Shopify and Shop Pay: catalog, payment, eligibility and attribution. An analysis of the official Meta and Shopify sources reviewed on September 9, 2026.
- Canonical HTML: https://verityscore.io/en/blog/muse-meta-shopify/
- Markdown alternate: https://verityscore.io/en/blog/muse-meta-shopify.md
- Language: en
- Content type: blog
- Published: 2026-09-09
- Updated: 2026-09-09
- Tags: shopify, muse, meta, ai-commerce, product-data
<div class="muse-review-content">


<p>For a Shopify merchant, the priority is to distinguish four states: the announcement of a channel, access to catalog data, the ability to pay, and the recording of an order. For an SEO or acquisition team, a fifth state matters: being able to link that order to the journey that produced it. A presence in a catalog provides none of these proofs on its own.</p>


<section class="muse-ecosystem" aria-labelledby="muse-ecosystem-title">
<h2 id="muse-ecosystem-title">The companies behind the Muse ecosystem</h2>
<div class="muse-logo-grid">
<div class="muse-company"><div class="muse-company-logo"><img src="/images/articles/muse-meta-shopify/meta-logo.png" width="116" height="50" alt="Meta"><img class="ecosystem-muse" src="/images/articles/muse-meta-shopify/muse-logo.svg" width="32" height="32" alt="Muse"></div><span>Muse</span></div>
<div class="muse-company"><div class="muse-company-logo"><img src="/images/articles/muse-meta-shopify/stripe-logo.svg" width="96" height="40" alt="Stripe"></div><span>Link</span></div>
<div class="muse-company"><div class="muse-company-logo"><img class="ecosystem-icon" src="/shopify-bag.svg" width="36" height="42" alt=""><strong>Shopify</strong></div><span>Shop Pay · announced as coming soon</span></div>
<div class="muse-company"><div class="muse-company-logo"><img class="ecosystem-icon" src="/images/articles/muse-meta-shopify/1password-logo.png" width="42" height="42" alt=""><strong>1Password</strong></div><span>Support announced</span></div>
</div>
<p class="muse-logo-source">Companies and services cited in Meta’s official announcement of September 8, 2026. <a href="#ref-1">Source [1]</a></p>
</section>

## What Muse changes in the shopping journey


<p>Meta describes Muse as a personal agent able to work over time, use a browser and continue a task after the app is closed. Its commercial relevance therefore goes beyond answering a product search. The agent can prepare an action, come back when a condition changes and ask for approval before the purchase. The announcement provides for the Muse app and WhatsApp as conversation interfaces. It places the initial rollout in the United States, on iOS, Android and the web. <a href="#ref-1" aria-label="Source 1">[1]</a></p>
<p>The technical documentation specifies the payment flow described at launch: Link provides a single-use card tied to a merchant, an amount and a validity period. The agent asks for human approval. Shop Pay also appears there as an upcoming integration. These are capabilities declared by Meta, with no purchase test carried out for this analysis. <a href="#ref-2" aria-label="Source 2">[2]</a></p>
<figure class="media-card">
<div class="media-label muse-brand-label"><img src="/images/articles/muse-meta-shopify/muse-logo.svg" width="36" height="36" alt="Official Muse by Meta logo"> Official demonstration · Muse by Meta</div>
<video poster="/images/articles/muse-meta-shopify/muse-video-poster.png" controls playsinline preload="none" aria-label="Muse shopping demonstration published by Meta" width="960" height="836"><source src="https://about.fb.com/wp-content/uploads/2026/09/Muse_Shopping.mp4" type="video/mp4"></video>
<figcaption>Meta illustrates a shopping scenario in its announcement of September 8, 2026. This demonstration by the publisher does not constitute a purchase tested for this article. The text above describes the documented journey. <a href="https://about.fb.com/news/2026/09/introducing-muse-personal-ai-agent/">Source and context</a> · <a href="https://about.fb.com/wp-content/uploads/2026/09/Muse_Shopping.mp4">Open the video</a>.</figcaption>
</figure>

<aside class="social-card" aria-label="Muse launch post">
<div class="media-label">At the source · post on X</div>
<p>The Muse account announces its launch on September 8, 2026. This post establishes the launch communication, without proving access or payment for any given store.</p>
<iframe class="muse-tweet" src="https://platform.twitter.com/embed/Tweet.html?id=2097399178376671666&amp;dnt=true&amp;lang=en&amp;theme=light&amp;width=550px" title="Official Muse post on X" referrerpolicy="no-referrer" loading="eager" width="550" height="780"></iframe><a class="source-link" href="https://x.com/Muse/status/2097399178376671666" rel="nofollow noopener noreferrer" target="_blank">View the post on X ↗</a></aside>

<p>Stripe provides a useful cross-check in its press release of September 8: US consumers can connect Link to Muse. It describes payment with the saved payment method at merchants that accept Link, and a single-use virtual card for other supported purchases. The amount must be approved in the conversation. This confirmation covers Link, without establishing the availability of Shop Pay in Muse. <a href="#ref-17" aria-label="Source 17">[17]</a></p>
<p>The implication for a store is concrete: an agent that browses may encounter the product page, the shipping terms and the site&#39;s checkout. A channel that offers an integrated checkout may present a different interface. Data, customizations and measurement do not necessarily follow the same path. The journey must be identified before concluding that a catalog, theme or tracking change will solve the problem.</p>
<div class="scroll" tabindex="0" role="region" aria-label="Table, horizontal scrolling available"><table class="states">
<thead>
<tr>
<th>State</th>
<th>What can be established</th>
<th>What remains to be proven</th>
</tr>
</thead>
<tbody><tr>
<td>Announcement</td>
<td>A platform describes a capability and a timeline</td>
<td>Access for a given user</td>
</tr>
<tr>
<td>Catalog</td>
<td>Data is accessible to a channel</td>
<td>Its selection and display for a query</td>
</tr>
<tr>
<td>Payment</td>
<td>A flow is documented for an eligible scope</td>
<td>Whether it works for the product, the country and the store</td>
</tr>
<tr>
<td>Order</td>
<td>An order appears in the merchant system</td>
<td>Its precise origin, its payment and its outcome after returns</td>
</tr>
</tbody></table></div>
<figure class="muse-diagram"><div class="diagram-heading"><span class="diagram-index">01</span><h3>Four states, four proofs</h3></div><div class="diagram-grid"><div class="diagram-card"><span class="diagram-step">01</span><h4>Announcement</h4><p>A channel is presented by the platform.</p></div><div class="diagram-card"><span class="diagram-step">02</span><h4>Catalog</h4><p>Product data is accessible.</p></div><div class="diagram-card"><span class="diagram-step">03</span><h4>Payment</h4><p>A payment flow is available.</p></div><div class="diagram-card"><span class="diagram-step">04</span><h4>Order</h4><p>A purchase is actually recorded.</p></div></div><figcaption>An announcement proves neither merchant access, nor payment, nor an order. A reading framework, with no performance data.<a class="figure-brand" href="/en/blog/"><img src="/logo.svg" width="20" height="20" alt="">Articles by Verity Score</a></figcaption></figure>
<p>This grid is a reading method, not a measure of Muse&#39;s performance. In particular, it avoids turning “available products” into “completed sales”.</p>


## Shop Pay: a discrepancy that requires naming the surface


<p>Meta&#39;s page of September 8 explicitly states that Shop Pay is coming soon. The technical publication on safety repeats this caveat. In parallel, the Shopify help page “Selling on Meta” describes a direct checkout enabled by default for eligible stores. It refers to Meta surfaces, without establishing a timeline specific to Muse and Shop Pay. <a href="#ref-1" aria-label="Source 1">[1]</a><a href="#ref-2" aria-label="Source 2">[2]</a><a href="#ref-3" aria-label="Source 3">[3]</a></p>
<figure class="source-capture">
<a href="https://about.fb.com/news/2026/09/introducing-muse-personal-ai-agent/"><img src="/images/articles/muse-meta-shopify/capture-meta-shop-pay.png" width="215" height="27" alt="Exact excerpt from Meta: Shop Pay is coming soon"></a>
<figcaption>Actual capture of the passage from Meta’s announcement, cropped to the Shop Pay mention, accessed September 9, 2026. The passage reads: Shop Pay is coming soon. <a href="https://about.fb.com/news/2026/09/introducing-muse-personal-ai-agent/">Read the full paragraph at Meta</a>.</figcaption>
</figure>

<figure class="muse-diagram"><div class="diagram-heading"><span class="diagram-index">02</span><h3>Shop Pay: two scopes to keep apart</h3></div><div class="diagram-grid"><div class="diagram-card"><span class="diagram-step">01</span><h4><img class="brand-muse" src="/images/articles/muse-meta-shopify/muse-logo.svg" width="28" height="28" alt="Muse by Meta logo"> Meta · Muse</h4><p>Meta’s announcement states that Shop Pay is coming soon.</p></div><div class="diagram-card"><span class="diagram-step">02</span><h4><img class="brand-shopify" src="/shopify-bag.svg" width="24" height="28" alt=""> Shopify · Meta surfaces</h4><p>The Help Center documents a direct checkout for eligible stores.</p></div></div><figcaption>The two sources do not describe the same scope. No general timeline for Shop Pay in Muse can be inferred from them.<a class="figure-brand" href="/en/blog/"><img src="/logo.svg" width="20" height="20" alt="">Articles by Verity Score</a></figcaption></figure>
<p>A difference in timeline, journey or population could explain the gap. These are hypotheses. The sources reviewed do not make it possible to choose between them. The defensible wording is therefore the following: <strong>Shopify documents Meta direct checkout; the general availability of Shop Pay in Muse is not confirmed by the Meta announcement reviewed.</strong></p>
<div class="scroll" tabindex="0" role="region" aria-label="Table, horizontal scrolling available"><table>
<thead>
<tr>
<th>Dated milestone</th>
<th>Source</th>
<th>Actual scope</th>
</tr>
</thead>
<tbody><tr>
<td>January 11, 2026</td>
<td>Shopify announces UCP, new integrations and the Agentic plan</td>
<td>Infrastructure context; no proof of the Muse launch</td>
</tr>
<tr>
<td>May 11, 2026</td>
<td>Shopify Changelog: new Agentic page in the admin</td>
<td>Centralization of settings and measurement</td>
</tr>
<tr>
<td>July 7, 2026</td>
<td>Displayed date of the Agentic Storefronts terms reviewed</td>
<td>Contractual framework; not the availability date of each channel</td>
</tr>
<tr>
<td>September 8, 2026</td>
<td>Meta presents Muse</td>
<td>US rollout, Link described, Shop Pay coming soon</td>
</tr>
<tr>
<td>September 9, 2026</td>
<td>Review of the Shopify and Meta documents</td>
<td>Documentary snapshot, with no test order</td>
</tr>
</tbody></table></div>
<p>Sources for the timeline: Shopify, announcements and terms; Meta, launch announcement. Documentation with no visible update date is not artificially dated to the day it was accessed. <a href="#ref-1" aria-label="Source 1">[1]</a><a href="#ref-8" aria-label="Source 8">[8]</a><a href="#ref-9" aria-label="Source 9">[9]</a><a href="#ref-10" aria-label="Source 10">[10]</a></p>


## Who is eligible, and for what?


<p>Shopify states that Meta direct checkout is shown to customers located in the United States. The stated condition is selling to US customers. This makes it possible neither to announce availability for French shoppers, nor to conclude that all merchants established in France are excluded. The shopper&#39;s country and the seller&#39;s country of establishment are two different pieces of information. <a href="#ref-3" aria-label="Source 3">[3]</a></p>
<p>Catalog eligibility is another filter. Shopify notably requires a Starter plan or higher, with a documented exception for the Agentic plan, a store that is not in private mode, a title, at least one image and a price greater than zero. Publication on compatible channels and an identifiable product URL also count. Products that are unlisted or hidden from search engines are excluded. Shopify also reviews the account&#39;s standing and history; a correctly completed product listing therefore does not guarantee the store&#39;s admission. <a href="#ref-4" aria-label="Source 4">[4]</a></p>
<div class="scroll" tabindex="0" role="region" aria-label="Table, horizontal scrolling available"><table>
<thead>
<tr>
<th>Situation</th>
<th>Documented state as of September 9</th>
<th>Consequence for the merchant</th>
</tr>
</thead>
<tbody><tr>
<td>Shopper in the United States</td>
<td>Muse rollout announced; Meta direct checkout planned for US customers</td>
<td>Verify access and the exact journey</td>
</tr>
<tr>
<td>Shopper in France</td>
<td>The announcement reviewed does not confirm a French launch</td>
<td>Do not present Muse as an available local channel</td>
</tr>
<tr>
<td>Eligible store with eligible products</td>
<td>Meta direct checkout described as active by default</td>
<td>Read the channel&#39;s actual state in the admin</td>
</tr>
<tr>
<td>Subscription, bundle, customizable or B2B-only product</td>
<td>Not supported in Meta direct checkout</td>
<td>Do not extend catalog eligibility to this checkout</td>
</tr>
<tr>
<td>Agentic plan without a main Shopify store</td>
<td>Specific procedure; direct checkout disabled by default</td>
<td>Do not apply the automations of the other plans</td>
</tr>
<tr>
<td>Meta shoppable ads</td>
<td>Facebook and Instagram channel with feed sync required</td>
<td>Distinguish this feed from discovery via Catalog</td>
</tr>
</tbody></table></div>
<p>The payment restrictions come from Shopify&#39;s Meta page, the catalog rules from its eligibility page, and the Agentic plan exception from its dedicated guide. This table is not a verdict on any particular store. <a href="#ref-3" aria-label="Source 3">[3]</a><a href="#ref-4" aria-label="Source 4">[4]</a><a href="#ref-7" aria-label="Source 7">[7]</a></p>
<p>The limits also affect the experience. Shopify excludes certain checkout blocks from Meta direct checkout, as well as local delivery, in-store pickup and pickup points. Automatic discounts and discount codes are described as supported. Shopify Functions relating to, among other things, shipping and validation may work, but this does not prove that every site customization carries over into Meta. <a href="#ref-3" aria-label="Source 3">[3]</a></p>


## Catalog: the integration reduces work, it does not erase the checks


<p>Shopify&#39;s page dedicated to Meta states that discovery relies on Shopify Catalog. It explicitly ties the Facebook and Instagram feed requirement to shoppable ads. Another page, dedicated to product discovery, more broadly describes Meta surfaces as powered by that Facebook and Instagram channel. The documentation is therefore not perfectly aligned in its wording. <a href="#ref-3" aria-label="Source 3">[3]</a><a href="#ref-5" aria-label="Source 5">[5]</a></p>
<p>To handle this discrepancy, the Meta-specific page is the most precise reference on the discovery/advertising distinction. It does not justify installing an advertising channel simply because an email mentions Muse. For a given store, the right check is to examine the intended surface, its state in Agentic and its current terms. An existing integration should not be modified on the basis of a generalization drawn from another page.</p>
<p>Shopify describes Catalog as a structured source of titles, descriptions, options, images, prices and availability, updated continuously. This promise does not provide a measured maximum delay between a stock change and its reflection in Muse. Nor does it demonstrate that the right variant will always be selected for an ambiguous request. <a href="#ref-5" aria-label="Source 5">[5]</a></p>
<p>An ecommerce team can prepare a verification without changing anything: choose a SKU and a variant, note their price and currency, record availability for the target market, then compare the information at each accessible step of the journey. If a Muse observation becomes possible, it must carry the same timestamp and the same context. A discrepancy should be kept as an observed discrepancy; its cause is not automatically a faulty Shopify sync.</p>
<p>For catalogs that use custom fields, the issue may be mapping: the data exists in the store but is not carried into the field expected by Catalog. Shopify documents this possibility. In SEO, the useful work then consists of making the attributes actually available and understandable, with the usage constraints, dimensions or compatibilities that make a decision possible. Adding adjectives or lengthening a description without information does not provide that proof. <a href="#ref-5" aria-label="Source 5">[5]</a></p>
<p>The Verity method on <a href="https://verityscore.io/en/blog/chatgpt-shopify-product-representation-gap/">product representation gaps</a> provides a complementary framework for comparing the approved fact with what is displayed. Its field is ChatGPT; its findings specific to that platform are not results measured on Muse. <a href="#ref-13" aria-label="Source 13">[13]</a></p>
<aside class="muse-reading"><span class="media-label">To prepare your catalog</span><p>To extend these checks, the <a href="https://universalcommerceprotocol.blog/catalogues-produits-pour-ia/">guide to product catalogs for AI agents</a> offers a reading grid for identifiers, attributes, prices, stock and shipping or return policies. It is a methodological complement; it does not demonstrate an integration of Muse with the UCP protocol.</p></aside>


## A Meta order does not measure Muse’s contribution on its own


<p>Shopify states that orders placed on Meta surfaces are attributed to Meta in the admin. The documentation reviewed does not demonstrate the existence of a field that isolates Muse across all of these orders. An increase in the Meta channel therefore cannot automatically become an increase in Muse sales. <a href="#ref-3" aria-label="Source 3">[3]</a></p>
<p>Direct checkout introduces another measurement break. Shopify specifies that Google Analytics and custom pixels do not fire in that checkout. The events described are sent server to server for the start and the end of checkout. The acquisition lead should avoid comparing this journey, without precaution, to a web conversion measured by browser-side pixels. <a href="#ref-3" aria-label="Source 3">[3]</a></p>
<p>The Agentic page presents reports on sales, orders, store sessions and conversions from those sessions. These measures do not all share the same scope. Dividing all orders attributed to a channel by store sessions alone can produce a meaningless rate if some orders were completed off-site. <a href="#ref-6" aria-label="Source 6">[6]</a></p>
<figure class="muse-diagram"><div class="diagram-heading"><span class="diagram-index">03</span><h3>From visibility to order</h3></div><div class="diagram-grid"><div class="diagram-card"><span class="diagram-step">01</span><h4>Visibility</h4><p>The product appears in an answer or a catalog.</p></div><div class="diagram-card"><span class="diagram-step">02</span><h4>Visit</h4><p>A session is measured on the store.</p></div><div class="diagram-card"><span class="diagram-step">03</span><h4>Attributed order</h4><p>A sale is linked to a source based on the available data.</p></div></div><figcaption>These measures answer different questions. No Muse traffic, conversion or revenue figure is represented.<a class="figure-brand" href="/en/blog/"><img src="/logo.svg" width="20" height="20" alt="">Articles by Verity Score</a></figcaption></figure>
<p>For each analysis, keep the period, the available channel or referrer, the known place of completion, and the cancelled or refunded orders. If the Muse origin is not exposed, the category remains “Meta, surface not isolated”. This discipline leaves an unknown visible rather than fabricating a result.</p>
<p>The <a href="https://verityscore.io/en/blog/share-of-ai-voice-guide-2026/">Verity guide to Share of AI Voice</a> is useful for defining visibility as a distinct indicator. A presence in an answer, a visit and an order have different denominators. No Muse traffic volume, conversion rate or revenue was measured in this analysis. <a href="#ref-14" aria-label="Source 14">[14]</a></p>


## Fees, returns and data: what the merchant keeps


<p>Shopify&#39;s Meta page states that no specific fees are added for selling through its direct checkout, apart from the usual payment processing fees. “No additional fees” therefore does not mean “free transaction”. The Agentic Storefronts terms also provide for the possibility of fees depending on the channel, presented in the admin. The current rule documented for Meta does not constitute a guarantee of permanent free use for all channels. <a href="#ref-3" aria-label="Source 3">[3]</a><a href="#ref-8" aria-label="Source 8">[8]</a></p>
<p>The merchant remains the merchant of record. The merchant manages orders, their fulfillment, returns and customer service. A purchase initiated in a conversation does not automatically transfer these responsibilities to the agent. The store&#39;s policies must be present and up to date; they are part of the terms published by Shopify. <a href="#ref-3" aria-label="Source 3">[3]</a></p>
<p>Meta also mentions Link protections. The Link page reviewed reserves them for eligible purchases and specifies that Cover Genius assesses coverage. It would be wrong to promise them for every Shopify order or for a future Shop Pay checkout. Customer service must first identify the payment method, the journey and the terms applicable to the case. <a href="#ref-1" aria-label="Source 1">[1]</a><a href="#ref-11" aria-label="Source 11">[11]</a></p>
<p>On the data side, Shopify describes channel access to product information and, when an order is placed through a direct checkout, to the necessary data such as the customer&#39;s name, email or address. The documentation states that the full order history and customer base are not opened to the channel. This remains different from an absence of sharing: the data needed for the transaction does flow. <a href="#ref-12" aria-label="Source 12">[12]</a></p>


## Muse’s privacy limits are already documented


<p>Meta describes a virtual machine dedicated to each user, isolated credential storage and a Sentinel agent that controls outbound actions. The same publication acknowledges that Muse can make mistakes and be subject to prompt injection attacks. These descriptions constitute the architecture declared by its publisher; they do not amount to independent certification in this analysis. <a href="#ref-2" aria-label="Source 2">[2]</a></p>
<p>The distinction between the virtual machine versions is essential. Meta states that the current architecture may allow access to certain data in order to operate, secure or support the service. The Confidential VM version, meant to cryptographically prevent that access, is announced for later in the year. Presenting this future property as achieved at launch would be inaccurate. <a href="#ref-2" aria-label="Source 2">[2]</a></p>
<p>The advertising policy calls for the same precision. Meta says it does not pass conversations or VM data to its advertising systems. But its documentation specifies that visits made by Muse to websites may be treated as the user&#39;s activity and influence the ads they see afterward. The absence of direct sharing of conversations does not mean the absence of any advertising consequence of the journey. <a href="#ref-2" aria-label="Source 2">[2]</a></p>
<p>Finally, Meta explains that it uses interaction trajectories scrubbed of certain personal information to train its models, with an opt-out option in the settings. This mechanism must be distinguished from credential storage, advertising sharing and the order data accessible to partners. These are different processing operations, subject to different controls. <a href="#ref-2" aria-label="Source 2">[2]</a></p>


## The five decisions to make before increasing acquisition effort


<div class="scroll" tabindex="0" role="region" aria-label="Table, horizontal scrolling available"><table>
<thead>
<tr>
<th>Proposed priority</th>
<th>Owner</th>
<th>Useful verification</th>
<th>Decision the evidence allows</th>
</tr>
</thead>
<tbody><tr>
<td>Qualify the scope</td>
<td>Ecommerce</td>
<td>Shopper country, product, channel and enabled payment</td>
<td>Identify a candidate journey, without announcing universal availability</td>
</tr>
<tr>
<td>Check one SKU</td>
<td>Catalog and SEO</td>
<td>Variant, price, currency, stock, attributes and policies</td>
<td>Correct an identified missing or contradictory data point</td>
</tr>
<tr>
<td>Examine the exceptions</td>
<td>Ecommerce and technical</td>
<td>Bundles, customization, subscriptions, shipping, checkout functions</td>
<td>Avoid a promise that the direct journey cannot keep</td>
</tr>
<tr>
<td>Define the measurement</td>
<td>Acquisition and analytics</td>
<td>Available attribution, place of payment, sessions, cancellations and returns</td>
<td>Distinguish channel observation from incremental revenue</td>
</tr>
<tr>
<td>Review the controls</td>
<td>Store manager</td>
<td>Catalog, direct checkout, future enrollment and terms</td>
<td>Choose channels with knowledge of their effects</td>
</tr>
</tbody></table></div>
<p>This plan is a recommendation. It starts with reversible readings and comparisons. It does not set a universal timeline or an advertising budget: those depend on the catalog, on access to the product and on the results actually observed.</p>
<p>The settings are found under Sales channels &gt; Agentic. Shopify allows catalog access and direct checkout to be managed separately when automatic management is disabled. Removing Catalog can take up to seven days for sharing to stop, according to its documentation, and does not necessarily prevent discovery through web crawling. Hiding a product entirely may affect its visibility in search engines; this is not a simple Muse preference. <a href="#ref-6" aria-label="Source 6">[6]</a></p>
<p>The launch justifies preparing the data and the measurement. It does not yet justify presenting Muse as a proven revenue source for a store. The next useful result is a verifiable trace linking a product, an eligible shopper, a payment flow and an attributable order. As long as one of these elements is missing, it must remain named.</p>


## Sources and methodology


<p>Documentary analysis closed on September 9, 2026. Official sources searched and then re-read, claims cross-checked by surface and date. No store, messaging or customer data consulted; no purchase, setting change or field test. Hypotheses and recommendations are distinguished from the platforms&#39; statements. Local appendices retain the extractions, contradictions and checks.</p>
<p id="ref-1"><a href="#ref-1" aria-label="Source 1">[1]</a> Meta. <a rel="nofollow noopener noreferrer" href="https://about.fb.com/news/2026/09/introducing-muse-personal-ai-agent/">Introducing Muse: The World’s First Personal AI Agent Built for Everyone</a>. September 8, 2026. Rollout announcement, not proof of universal access.</p>
<p id="ref-2"><a href="#ref-2" aria-label="Source 2">[2]</a> Meta. <a rel="nofollow noopener noreferrer" href="https://security.muse.ai/">How We Built Safety Into Muse</a>. September 8, 2026 according to the metadata; accessed September 9. Architecture, purchases, data and limits declared by the publisher.</p>
<p id="ref-3"><a href="#ref-3" aria-label="Source 3">[3]</a> Shopify Help Center. <a rel="nofollow noopener noreferrer" href="https://help.shopify.com/en/manual/online-sales-channels/agentic-storefronts/meta">Selling on Meta</a>. Accessed September 9, 2026; update date not displayed in the body reviewed. Meta scope, with no Muse-specific timeline.</p>
<p id="ref-4"><a href="#ref-4" aria-label="Source 4">[4]</a> Shopify Help Center. <a rel="nofollow noopener noreferrer" href="https://help.shopify.com/en/manual/shopify-catalog/requirements">Requirements for being included in Shopify Catalog</a>. Accessed September 9, 2026. Catalog conditions, not a guarantee of recommendation.</p>
<p id="ref-5"><a href="#ref-5" aria-label="Source 5">[5]</a> Shopify Help Center. <a rel="nofollow noopener noreferrer" href="https://help.shopify.com/en/manual/online-sales-channels/agentic-storefronts/products">Shopify Catalog and product discovery for agentic storefronts</a>. Accessed September 9, 2026. Wording discrepancy on Meta surfaces retained.</p>
<p id="ref-6"><a href="#ref-6" aria-label="Source 6">[6]</a> Shopify Help Center. <a rel="nofollow noopener noreferrer" href="https://help.shopify.com/en/manual/online-sales-channels/agentic-storefronts/agentic-home">Managing agentic storefronts</a>. Accessed September 9, 2026. Settings and reports described, without verification of an account.</p>
<p id="ref-7"><a href="#ref-7" aria-label="Source 7">[7]</a> Shopify Help Center. <a rel="nofollow noopener noreferrer" href="https://help.shopify.com/en/manual/online-sales-channels/agentic-storefronts/agentic-plan-setup">Setting up agentic storefronts for stores on the Agentic plan</a>. Accessed September 9, 2026. Procedure for a separate plan.</p>
<p id="ref-8"><a href="#ref-8" aria-label="Source 8">[8]</a> Shopify. <a rel="nofollow noopener noreferrer" href="https://www.shopify.com/legal/terms-agentic-storefronts">Agentic Storefronts Supplemental Terms of Service</a>. Displayed update: July 7, 2026, fresh read on September 9. An earlier extraction returning May 25 was discarded for the date and version.</p>
<p id="ref-9"><a href="#ref-9" aria-label="Source 9">[9]</a> Shopify. <a rel="nofollow noopener noreferrer" href="https://www.shopify.com/news/ai-commerce-at-scale">The agentic commerce platform</a>. January 11, 2026. Historical context, without projecting UCP capabilities onto Muse.</p>
<p id="ref-10"><a href="#ref-10" aria-label="Source 10">[10]</a> Shopify Changelog. <a rel="nofollow noopener noreferrer" href="https://changelog.shopify.com/posts/a-new-home-for-agentic-storefronts-in-your-admin">A new home for Agentic Storefronts in your admin</a>. May 11, 2026.</p>
<p id="ref-11"><a href="#ref-11" aria-label="Source 11">[11]</a> Link. <a rel="nofollow noopener noreferrer" href="https://support.link.com/questions/what-s-covered-with-protections">What’s covered with protections?</a>. Accessed September 9, 2026. Conditional coverage; no individual contract verified.</p>
<p id="ref-12"><a href="#ref-12" aria-label="Source 12">[12]</a> Shopify Help Center. <a rel="nofollow noopener noreferrer" href="https://help.shopify.com/en/manual/online-sales-channels/agentic-storefronts/data-privacy">Data sharing and privacy for selling with agentic storefronts</a>. Accessed September 9, 2026.</p>
<p id="ref-13"><a href="#ref-13" aria-label="Source 13">[13]</a> Verity Score. <a rel="nofollow noopener noreferrer" href="https://verityscore.io/en/blog/chatgpt-shopify-product-representation-gap/">ChatGPT and Shopify: Audit the Product Representation Gap</a>. August 31, 2026; re-read September 9. Methodological complement, not Muse proof.</p>
<p id="ref-14"><a href="#ref-14" aria-label="Source 14">[14]</a> Verity Score. <a rel="nofollow noopener noreferrer" href="https://verityscore.io/en/blog/share-of-ai-voice-guide-2026/">Share of AI Voice: 5 Engines, Not 1</a>. Accessed September 9, 2026. Definitional complement, without reuse of its metrics.</p>
<p id="ref-15"><a href="#ref-15" aria-label="Source 15">[15]</a> Meta. <a rel="nofollow noopener noreferrer" href="https://introducing.muse.ai/">How We Designed Muse</a>. September 2026. Design reference read for context; team anecdotes not reused as independent observations.</p>
<p id="ref-16"><a href="#ref-16" aria-label="Source 16">[16]</a> Shopify. <a rel="nofollow noopener noreferrer" href="https://www.shopify.com/agentic-storefronts">Sell in AI chats with Shopify</a>. Accessed September 9, 2026. Commercial page checked against the more precise terms of the Help Center.</p>

<p id="ref-17"><a href="#ref-17" aria-label="Source 17">[17]</a> Stripe. <a rel="nofollow noopener noreferrer" href="https://stripe.com/en-ca/newsroom/news/stripe-helps-meta-muse-shop-with-link">Stripe helps Muse shop across the internet with Link</a>. September 8, 2026; accessed September 9. Press release from the payment platform, with no independent purchase.</p>

<p id="ref-18">[18] Universal Commerce Protocol Blog. <a href="https://universalcommerceprotocol.blog/catalogues-produits-pour-ia/">Guide des catalogues produits pour les agents IA</a>. Accessed September 9, 2026. Methodological complement on catalog structure; does not prove a UCP integration in Muse.</p>


</div>