# Claims & Proof: How AI Agents Verify Your Product Claims > AI agents fact-check your marketing claims before recommending products. Learn how claims detection, proof verification, and credibility scoring work. - Canonical HTML: https://verityscore.io/en/kb/claims-proof/ - Markdown alternate: https://verityscore.io/en/kb/claims-proof.md - Language: en - Content type: kb - Published: 2026-02-07 - Updated: 2026-04-19 - Tags: claims, proof, credibility, anti-hallucination - Audit zone: 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**. In practice, 96% of Google AI Overview citations come from sources with strong E-E-A-T signals ([Wellows, 2026](https://wellows.com/blog/google-ai-overviews-ranking-factors/)). The foundational research in the field confirms that adding authoritative citations boosts AI visibility by up to +115%, and that direct quotations bring an additional +28% visibility ([Princeton GEO, KDD 2024](https://arxiv.org/abs/2311.09735)). 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`](https://schema.org/Offer) schema with `shippingDetails`. Brand mentions correlate at r=0.664 with AI citation probability ([SurferSEO, 2025](https://surferseo.com/blog/ai-citation-report/)). Security is also part of this verification: Microsoft Security documented over 50 prompt injection attempts in February 2026 aimed at poisoning AI recommendations, which pushes engines to favor claims backed by structured data ([Microsoft Security, 2026](https://www.microsoft.com/en-us/security/blog/2026/02/10/ai-recommendation-poisoning/); [OWASP LLM01:2025](https://owasp.org/www-project-top-10-for-large-language-model-applications/)). ## What AI actually checks
AI agent claim verification loop: claim detected, structured proof verification, verdict - proven claim (trust) or unproven claim (distrust, product ignored)
Figure 1 - How AI agents verify your marketing claims
For each type of claim, AI agents look for a corresponding structured proof: policy pages for shipping/return promises, [`AggregateRating`](https://schema.org/AggregateRating) for review claims, [`Offer`](https://schema.org/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 --- ## Related articles - [E-E-A-T Signals for AI: How Agents Evaluate Trust](/en/kb/eeat-signals-ai) - [Schema.org Product: Why and How on Shopify](/en/kb/schema-org) - [AggregateRating: Making Your Reviews AI-Readable](/en/kb/aggregate-rating) - [Conversational Content: Writing for Humans AND AI](/en/kb/conversational-content) - [GEO Audit: The Complete Guide to Optimizing Your Shopify Store for AI](/en/kb/geo-audit) --- **Ready to check your store?** [Run a free GEO audit →](https://verityscore.io) ## FAQ ### What is a marketing claim vs a proof in e-commerce? A claim is what you display to visitors ('Free shipping in 24h', '4.8/5 rating'). A proof is verifiable structured data that confirms it, like a shippingDetails schema or an AggregateRating in JSON-LD. AI trusts claims only when proofs exist. ### How do AI agents verify product claims on my Shopify store? AI agents cross-check visible claims against structured data in your HTML. If your banner says 'Free shipping' but there's no shippingDetails in your schema.org, or your policy page is a 404, the AI treats the claim as unverified and less trustworthy. ### Why does claim-proof coherence matter for AI visibility? LLMs are trained to avoid hallucination, so they prefer sources with verifiable information. A store where every marketing promise is backed by structured data gets recommended more often than one with unverifiable claims. ### What are the most common claim-proof gaps on Shopify stores? The most frequent gaps are: review ratings displayed as images but absent from schema, shipping promises without a shippingDetails schema, return guarantees without an accessible policy page, and price claims that don't match the Offer schema. ## Sources - [Google AI Overviews Ranking Factors: 2026 Guide (Wellows/StackMatix)](https://wellows.com/blog/google-ai-overviews-ranking-factors/) (industry) - [AI Citation Report 2025: Sources AI Overviews Trust Most](https://surferseo.com/blog/ai-citation-report/) (industry) - [GEO: Generative Engine Optimization (Princeton, KDD 2024)](https://arxiv.org/abs/2311.09735) (academic) - [Microsoft Security: AI Recommendation Poisoning (February 2026)](https://www.microsoft.com/en-us/security/blog/2026/02/10/ai-recommendation-poisoning/) (official) - [OWASP Top 10 for Large Language Model Applications (LLM01:2025)](https://owasp.org/www-project-top-10-for-large-language-model-applications/) (official) - [Schema.org Offer Specification](https://schema.org/Offer) (official) - [Schema.org AggregateRating Specification](https://schema.org/AggregateRating) (official)