Why does Product schema decide AI shopping visibility?
When someone asks ChatGPT or Gemini ‘best mechanical keyboard under $150’, the engine retrieves product pages and listings, grounds the facts — price, availability, rating, specs — across sources, then generates a recommendation with citations. Product schema gives that pipeline labeled fields instead of prose to parse. Since there are no ad slots in ChatGPT answers as of November 2026, placement cannot be bought; it is earned by being the most reliably parseable, verifiable option. Full channel mechanics in ChatGPT shopping.
Which fields matter most?
- name, brand, description, image, sku, gtin. Identity fields that let engines reconcile your product with the same product on retailers and review sites. GTINs are powerful entity keys.
- offers (Offer): price, priceCurrency, availability, url, priceValidUntil. The grounding-critical set. Stale or inconsistent prices are the fastest way to be dropped from recommendations.
- aggregateRating and review. Recommendation answers lean heavily on ratings — but they must reflect real, visible reviews; see Review and AggregateRating schema.
- category, additionalProperty. Specs (materials, dimensions, compatibility) that answer comparison sub-queries produced by query fan-out.
A minimal Product skeleton
{"@context":"https://schema.org","@type":"Product","@id":"https://example.com/p/ergochair#product","name":"ErgoChair Pro","brand":{"@id":"https://example.com/#org"},"sku":"ECP-100","gtin13":"0123456789012","image":["https://example.com/img/ecp.jpg"],"description":"...","offers":{"@type":"Offer","url":"https://example.com/p/ergochair","priceCurrency":"USD","price":"349.00","availability":"https://schema.org/InStock"},"aggregateRating":{"@type":"AggregateRating","ratingValue":"4.6","reviewCount":"312"}}
How do consistency rules work across sources?
The grounding stage compares your markup against your visible page, your retailer listings, marketplace feeds and review platforms. The discipline that survives grounding:
- One source of truth: generate page, markup and merchant feeds from the same product database.
- Update markup the moment price or availability changes; pair with IndexNow so the Bing index — which ChatGPT search depends on — refreshes fast (IndexNow setup).
- Keep ratings real and reviewable on-page; invented ratings fail grounding and violate platform policies — see review management.
Common pitfalls
The classic failure is price drift: markup says $349, page says $379, marketplace says $365 — the model either hedges or excludes you. Second, marking up category pages as a single Product, which Google explicitly disallows and which confuses entity reconciliation. Third, availability mismatches during sales events, precisely when shopping-answer volume peaks. Fourth, hiding variants: if colors or sizes differ in price, model them with proper variant structure rather than one optimistic offer. Finally, shipping markup via tag manager where non-rendering AI crawlers never see it — product JSON-LD belongs in the initial HTML (JavaScript rendering and AI). Marketplaces face extra entity duplication issues; see GEO for marketplaces.
How do variants, bundles and marketplaces change the markup?
Variant handling is where product markup quietly fails. If size or color changes price, each purchasable variant needs its own offer with its own price and availability — a single optimistic ‘from $99’ offer mismatches reality the moment a model or a user picks a specific configuration. For marketplaces, the same product appears on many seller pages: keep GTINs identical everywhere so engines reconcile the entity, and make your page the most complete, most current version of the truth — that is what wins grounding tie-breaks; see GEO for marketplaces. Bundles should be marked up as their own products with components described in copy, not as fake discounts on individual SKUs.
Where do reviews and comparisons fit?
Shopping answers are comparative by nature: the fan-out behind ‘best X under Y’ includes review roundups and versus-queries, not just product pages. Pair strong product markup with genuinely comparative editorial content — see comparison content — and a real review footprint per Review schema. The product page proves the facts; the comparison and review corpus proves the claim that your product belongs in the recommendation.
Where to go next
Add ratings discipline in Review schema, nest products into your graph with JSON-LD nesting, and tune pricing pages via pricing pages for AI.