Skip to main content
Available on Shopify The Shopify app that audits and fixes what AI reads about your products. Verity Score is on Shopify Free trial
Critical Agentic Commerce

Get Shopify Products Into ChatGPT Shopping

Updated Recently updated
#chatgpt#chatgpt-shopping#agentic-storefronts#shopify#ai-commerce#agentic-commerce
Share

ChatGPT Shopping for Shopify: the documented path

Shopify documents ChatGPT as a discovery-focused referral channel. Eligible products can be shared through Shopify Catalog, and the buyer completes the purchase on the merchant’s own checkout. The store must sell to customers in the United States, although it can be based elsewhere.

Eligibility is not the same as retrieval, accurate representation or recommendation. Shopify and OpenAI do not publish a ChatGPT product-ranking formula. The merchant can improve the product record and verify the result; the external surface still controls selection.

For the whole-store sequence across Search, product pages, Catalog, feeds and measurement, follow the 27-check Shopify AI SEO guide.

How ChatGPT Recommends Your Products

Understanding ChatGPT’s recommendation mechanism is essential to optimizing for it. Here is the pipeline, step by step.

ChatGPT Shopping recommendation pipeline in 5 steps: user query, schema.org catalog interrogation, structured data relevance ranking, comparative recommendation, checkout on Shopify store via Agentic Storefronts
Figure 1 - How ChatGPT selects and recommends your Shopify products

Step 1 - The user query. A user asks ChatGPT something like “What is the best natural anti-wrinkle serum for sensitive skin, shipped to the US?” or “Compare the best carry-on suitcases under $150.”

Step 2 - Catalog interrogation. ChatGPT can rely on multiple sources: Shopify Catalog, merchant feeds, structured data, accessible pages, and third-party information. Schema.org matters, but it is not the only signal.

Step 3 - Relevance ranking. ChatGPT selects products based on user intent and the quality of available signals. A product with price, availability, GTIN, AggregateRating, specifications, shipping conditions, and return policy gives the agent stronger evidence than a product with only a name and price, without guaranteeing ranking.

Step 4 - Recommendation formulation. ChatGPT formulates a comparative response citing the most relevant products. It includes prices, ratings, differentiating features, and - via Agentic Storefronts - a direct link to your store for purchase.

Step 5 - Checkout on your store. The buyer clicks the link and lands on your Shopify store. They complete the purchase through your regular checkout, with your upsells, loyalty program, and branding.

The merchant-controlled part is the product record. Factual wording such as composition, use, price, availability, reviews, shipping and returns gives an external system more verifiable information than an unsupported superlative. Schema.org can help express some of those facts, but it is one surface among Catalog data, feeds, visible pages and third-party information.

Shopify Catalog and Agentic Surfaces: What to Verify

Shopify is deploying agentic surfaces that make catalogs more usable by AI assistants, especially through Shopify Catalog and Storefront Catalog MCP. But being on Shopify is not enough: ChatGPT Shopping, Copilot, Google AI Mode, Gemini, or Perplexity use different signals depending on country, channel, and merchant eligibility.

How to Verify Public Signals

To verify that your store is truly usable by AI agents:

  1. Test public product pages, schema.org, feeds, and policies.
  2. Verify Storefront Catalog MCP at https://your-store.com/api/ucp/mcp when available.
  3. Check that search_catalog, lookup_catalog, and get_product return complete and consistent products.
  4. Within the Shop channel, verify that Agentic Storefronts is enabled (green)
  5. Optionally, check in Settings > Checkout that your checkout is up to date

Which AI Agents Have Access

With Agentic Storefronts enabled, your catalog is accessible to:

  • ChatGPT (OpenAI) - via ChatGPT Shopping
  • Microsoft Copilot - via Bing Shopping integration
  • Google AI Mode - via Google Shopping Graph
  • Gemini (Google) - via Google Shopping Graph
  • Perplexity - via direct crawl and Shopping API

What It Costs

Appearing is free. Agentic Storefronts are included in all Shopify plans (Basic, Shopify, Advanced, Plus): no activation fee, and discovery in AI answers costs nothing.

The buyer is sent to your own Shopify checkout, so you stay the merchant of record: it is your store that processes the payment, not ChatGPT, and you pay only your regular Shopify fees. Shopify’s ChatGPT agentic storefront documentation states it plainly, that there are no fees associated with selling in ChatGPT and that you pay only your standard payment processing fees, with the purchase completed on your store checkout in a ChatGPT in-app browser or a new tab. In short, the AI agent is a qualified traffic source, not a paid intermediary.

There was an exception, and it is worth knowing because the figure still circulates. In January 2026 Shopify confirmed to The Information that its merchants would pay OpenAI 4% on sales closed through the in-chat checkout. That checkout was shut down in March 2026, OpenAI stating the first version lacked the flexibility it wanted (CNBC, March 24, 2026). The fee went with it. Any 2026 guide still quoting 4% as a live cost is describing a product that no longer exists.

Eight observable checks, not published ranking factors

The checks below are merchant diagnostics. They are not a ranking formula published by Shopify or OpenAI, and passing them cannot guarantee that a product is shown.

Criterion 1: Complete Product Schema

What works: a JSON-LD Product with name, description, brand (with name), image (direct URL), sku, gtin13, offers (price, priceCurrency, availability, url) and at least 5 additional properties (material, color, weight, etc.).

Weak evidence: a schema with only name and price leaves fewer facts for comparison and verification.

Criterion 2: AggregateRating

What works: an AggregateRating schema with ratingValue (e.g., 4.7), reviewCount (e.g., 234), and bestRating (5), injected into the Product JSON-LD and rendered in native HTML on the page.

What does not work: reviews only in a JavaScript widget (Judge.me in lazy-load mode, Loox iframe). The crawler reads the initial HTML, not the DOM after JavaScript execution.

Criterion 3: Structured Pricing

What works: an Offer with price as a numeric value (e.g., “29.90”), priceCurrency as ISO code (e.g., “USD”), and availability as a schema.org value (e.g., “https://schema.org/InStock”).

What does not work: a price only in the page text (“Starting from $29.90”) without a corresponding Offer schema. Or a price in the wrong currency (EUR for a US store).

Criterion 4: Structured Shipping

What works: a shippingDetails schema with shippingDestination (US), shippingRate (name + value + currency), and deliveryTime (handlingTime + transitTime in business days).

Weak evidence: shipping conditions that disagree between the policy page, product page, Catalog or structured data create ambiguity. No published ChatGPT rule says that only schema is read.

Criterion 5: Return Policy

What works: a hasMerchantReturnPolicy schema with returnPolicyCategory (e.g., MerchantReturnFiniteReturnWindow), merchantReturnDays (e.g., 30), returnFees (e.g., FreeReturn).

Weak evidence: a return policy hidden in generic footer copy or inconsistent across surfaces is harder to verify. The comparison value is a diagnostic priority, not a published ChatGPT ranking factor.

Criterion 6: Images with Alt Text

What works: high-resolution product images with descriptive alt texts (“Vitamin C serum 30ml - golden texture - amber glass bottle”) and a schema image with a direct URL (no JS lazy-load).

What does not work: images without alt text, or with generic alts (“product-image-1.jpg”). Multimodal models (GPT-4 Vision, Gemini) analyze images, and alt text provides them context.

Criterion 7: Real-Time Availability

What works: the availability field in the Offer schema updated in real time (InStock, OutOfStock, PreOrder). Shopify updates this field automatically if your schema is properly configured.

What does not work: a product showing “InStock” in the schema when it is actually out of stock. ChatGPT recommending an unavailable product = bad user experience = trust loss.

Criterion 8: GTIN/EAN

What works: a gtin13 (or gtin12, gtin14) field with the product’s real barcode. The GTIN allows ChatGPT to uniquely identify your product and cross-reference it with other sources (Amazon reviews, comparison sites, product databases).

What does not work: an empty GTIN field when a real identifier exists, or a fabricated GTIN. Use only identifiers assigned to the actual product.

Checklist: Optimize Your Store for ChatGPT

10 actions ordered by impact, with the exact Shopify Admin path.

  1. Verify Agentic Storefronts are active - Admin > Settings > Sales Channels > Shop > Agentic Storefronts. Confirm status is “Active.”

  2. Complete your Product schema - Theme > Edit code > product.liquid. Verify that the JSON-LD contains: name, description, brand, image, sku, gtin, offers (price, priceCurrency, availability), aggregateRating.

  3. Add GTINs/EANs to every product - Products > [Product] > Variants > Barcode. Fill in the real GTIN-13 or EAN. If you do not have a GTIN (artisan product), leave it empty rather than fabricating a code.

  4. Enable AggregateRating in native HTML - Review app settings > Schema markup / Rich snippets > Enable. Verify by “View Page Source” on a PDP that AggregateRating appears in the JSON-LD.

  5. Structure shipping conditions - Add shippingDetails to the JSON-LD schema of your products: destination, cost, delivery time, handling time.

  6. Structure your return policy - Add hasMerchantReturnPolicy to the schema: return window, fees, conditions. Create a /policies/refund-policy page if it does not exist.

  7. Optimize image alt texts - Products > [Product] > Images > click each image to add a descriptive alt text. Recommended format: “[Product name] - [visible detail] - [context].”

  8. Transform descriptions into factual content - Replace emotional copywriting with specifications: composition, dimensions, weight, certifications, proven results, factual comparisons.

  9. Verify real-time availability - Products > [Product] > Variants. Confirm that inventory tracking is enabled and quantities are current. The Offer schema must reflect actual availability.

  10. Segment robots.txt by AI bot role - Theme > Edit code > robots.txt.liquid. Do not block organic discovery agents such as OAI-SearchBot, ChatGPT-User, PerplexityBot, Claude-SearchBot, and Claude-User on public product and content pages. Decide separately on training crawlers such as GPTBot/ClaudeBot, usage controls such as Google-Extended, and paid-validation bots such as OAI-AdsBot.

How Verity Score Measures Your ChatGPT Readiness

Verity Score includes a specific agentic commerce readiness score: the ACP Readiness Score (Agentic Commerce Protocol), rated out of 100.

What the ACP Readiness Score Measures

The ACP Readiness Score evaluates your store’s ability to be recommended and sold by AI agents. It combines:

  • Product schema completeness: percentage of essential fields filled (name, price, gtin, brand, image, availability, rating, shipping, return policy)
  • Structured data quality: price consistency, GTIN validity, availability accuracy
  • HTML accessibility: is critical data in native HTML or only in JavaScript?
  • AI trust signals: AggregateRating present, verifiable evidence, consistent claims
  • Agentic protocols: Storefront Catalog MCP /api/ucp/mcp, canonical catalog tools, llms.txt file, schema/feed/page consistency

Sample Report

The Verity Score ACP Readiness report details each criterion with a status (present, absent, incomplete) and a precise corrective action. A preview:

CriterionStatusImpact
Product SchemaIncomplete (6/12 fields)High
AggregateRatingAbsent in HTMLCritical
GTIN/EANFilled on 40% of productsHigh
shippingDetailsMissing from schemaMedium
returnPolicyMissing from schemaMedium
robots.txtGPTBot blockedCritical
llms.txtMissingMedium
Image alt text70% coveredLow

How to interpret a re-test

Re-run the same source checks and a controlled set of shopping prompts after material corrections. A better product record proves that the merchant-controlled inputs improved; it does not prove that ChatGPT will retrieve or recommend the product. Record the date, market, prompt, cited sources and product facts so later observations are comparable.

Frequently Asked Questions

Is my Shopify store already visible on ChatGPT?
If your store is on Shopify and you have the Shop channel active, your products are technically accessible to ChatGPT via the Shopify catalog. However, being accessible does not mean being recommended. ChatGPT ranks products based on the quality of their structured data, and the majority of stores have insufficient data to appear in recommendations.
How much does selling through ChatGPT Shopping cost?
Nothing beyond what you already pay. Shopify's official documentation is explicit: there are no fees associated with selling in ChatGPT, and you pay only your standard payment processing fees. Buyers complete the purchase on your own store checkout, opened in a ChatGPT in-app browser or a new tab, so you stay the merchant of record. The 4% figure that circulated in January 2026 applied to OpenAI's in-chat checkout, which was shut down in March 2026.
Is ChatGPT Shopping available in my country?
Yes. ChatGPT Shopping is available to ChatGPT users in most countries, including the US, UK, Canada, Australia, and the EU. Recommendations adapt to the user's language and location. To maximize your visibility, ensure your structured data is complete in your target market's language.
How can I tell if my products appear in ChatGPT?
There is no official merchant console that reports every ChatGPT appearance. Use controlled purchase-related prompts as observations, then verify the product facts and eligibility at the source. Verity Score audits observable readiness signals; it does not calculate a recommendation probability.
Do I need Stripe to sell on ChatGPT?
No, not since March 2026. OpenAI ended Instant Checkout, its in-ChatGPT payment step, and buyers now complete the purchase on the merchant's own site. Product data, policies, feed readiness and checkout consistency are observable merchant inputs, but Shopify and OpenAI do not publish the formula used to select products.
What is the difference between Agentic Storefronts and Instant Checkout?
Instant Checkout was OpenAI's in-ChatGPT payment step; it was shut down in March 2026, with OpenAI stating the first version lacked the flexibility it wanted. Agentic Storefronts is Shopify's layer, still active, which exposes your catalog to ChatGPT and sends the buyer to your own checkout. Only the second one still exists, and product data quality is what determines agent confidence in it.