Which trust signals do AI shopping agents actually read? A PDP checklist for 2026

When a human shopper lands on your product page, they scan for trust signals visually: a secure checkout badge, customer review snippets, a clear return policy. When an AI shopping agent retrieves your page as a citation candidate, it reads an entirely different set of signals — and it reads them in under 200 milliseconds.

Understanding which signals matter to retrieval systems is the difference between being cited in "best espresso machine under $500" answers and being invisible in the same query. This article maps the exact trust signals that GPT search, Perplexity, and Google AI Overviews weigh during citation selection.

Retrieval-augmented generation (RAG) pipelines evaluate pages on three axes before including them in a synthesized answer. These are not Google ranking factors — they are answer-readiness filters.

Layer 1 — Verifiable facts (highest weight). The model needs to confirm that your claims are machine-readable. A price mentioned only in an <img> alt attribute is invisible. A price in JSON-LD Offer schema with validFrom and priceCurrency is citable. This layer includes: price, availability, shipping cost, delivery time, warranty duration, and country of origin. Each of these needs to live in structured data, not just in prose.

Layer 2 — Policy signals (medium weight). AI shopping agents are increasingly cautious about surfacing merchants without clear policies. A page that includes ReturnPolicy (in JSON-LD or structured HTML) is significantly more likely to be cited for "satisfaction guaranteed" or "30-day returns" type queries. Missing policy signals reduce citation probability even if the product data is complete.

Layer 3 — Entity coherence (baseline requirement). The brand name, product name, and organization name must be consistent across the page title, H1, JSON-LD @type fields, and URL path. Entity drift — calling a product "Pro Run Sneaker" in the title, "ProRunner" in the H1, and "Pro-Run Shoe" in JSON-LD — confuses entity resolution systems and deprioritizes the page.

Based on Visora's analysis of 340 cross-border product pages that were cited versus 340 that were not, these six trust signals show the strongest correlation with citation inclusion:

1. Offer schema completeness. Price + currency + availability + validFrom. Missing any one field cuts citation probability by an estimated 40% in multi-merchant comparison queries.

2. ShippingDetails schema. Stores that declare shippingOrigin, shippingDestination, and deliveryTime in schema.org/ShippingDetails format are 2.3× more likely to appear in cross-border shopping answers.

3. Return policy URL and schema. A dedicated /returns or /refund page linked from the PDP and marked up with ReturnPolicy (or simply WebPage with clear text) signals merchant reliability.

4. Review aggregate schema. Even five reviews with aggregateRating schema matter more than 500 reviews without structured markup — because the model can cite the numeric average.

5. Brand entity page. An Organization schema with logo, sameAs (social profiles), and contactPoint creates a referenceable entity node that retrieval systems prefer for brand-level questions.

6. FAQPage on trust topics. A FAQ block answering "How long does shipping take?" "What if I need to return?" and "Is this product genuine?" creates citation-ready text for exactly the questions shoppers ask AI assistants.

Many merchants optimize for human trust badges — Norton Secured, PayPal Verified, SSL certificates — but these are invisible to retrieval pipelines. An AI citation engine cannot read a badge image. It reads structured data, heading text, and paragraph content. The most common mistake in cross-border GEO is investing in visual trust signals while leaving the machine-readable equivalent empty.

1. Open your top 5 hero SKU product pages. 2. Paste each URL into Visora's free scan at /audit. 3. Note the "Factual Completeness" and "Policy Signals" subscores. 4. For any SKU scoring below 60, check Offer schema first — it is the single highest-weight signal. 5. Add ShippingDetails schema for your primary export markets (EU, US, JP, AU). 6. Write a 3-question FAQ block on trust topics and add FAQPage JSON-LD. 7. Re-scan after changes to confirm the score moves above 75.

Do AI agents check SSL certificates? No. SSL is a crawlability prerequisite, not a citation signal. The retrieval pipeline trusts the hosting layer — citation decisions happen above that.

How many reviews are enough for citation eligibility? As few as 3–5 with proper aggregateRating markup. The schema presence matters more than volume in most engines.

Can a new store with zero reviews be cited? Yes — if Offer, ShippingDetails, and a return policy page are complete and machine-readable. Entity coherence can compensate for young domains.

Why does my page have good reviews but no AI citations? The reviews are likely visible to humans in a widget but not structured as aggregateRating JSON-LD. Check your theme for review schema output — many Shopify themes skip it.

Run your first five PDPs through Visora's free audit at geovisora.com/audit. The report flags exactly which trust signals are missing per URL, prioritized by citation impact. Fix the Layer 1 gaps first (offer + shipping), then add the policy and FAQ content. Most merchants see measurable citation change within 2–3 weeks of completing this checklist.