Product Schema Markup

Product schema markup is structured data embedded in a product page using the Schema.org Product vocabulary, typically in JSON-LD format. It gives search engines and AI systems a machine-readable view of the product - name, description, brand, SKU, GTIN, price, availability, rating, shipping, and return policy - that does not depend on the engine correctly parsing the rendered HTML.

Schema.org was founded by Google, Microsoft, Yahoo, and Yandex in 2011 as a shared vocabulary for structured web data. The Product type has expanded steadily since, adding fields for offers, shipping details, merchant return policies, and - most recently - attributes relevant to AI shopping like additionalProperty for arbitrary key-value attributes.

JSON-LD is the preferred format because it is embedded as a self-contained script block in the page head rather than woven through the HTML. Google explicitly recommends JSON-LD over Microdata or RDFa. All major AI engines parse JSON-LD, which is why product schema is the single highest-leverage technical optimization for AI visibility.

A complete Product schema block for AI commerce includes:

Core identity fields (required):

• name - the product title
• image - one or more high-resolution product images (multiple angles preferred)
• description - a complete, differentiated product description
• brand - the manufacturer or brand, as a structured Brand object
• sku - your internal SKU

Commerce fields (required for shopping results):

• offers - containing price, priceCurrency, availability (InStock / OutOfStock / PreOrder), priceValidUntil, and url
• gtin / gtin13 / mpn - global product identifiers; required for Merchant Center and increasingly expected by AI engines

Trust and enrichment fields (strongly recommended):

• aggregateRating - with ratingValue, reviewCount, bestRating
• review - individual reviews with author, ratingValue, and reviewBody
• shippingDetails - shipping cost and delivery time
• hasMerchantReturnPolicy - return policy details per Schema.org's MerchantReturnPolicy type

Attribute depth (the differentiator):

• material, color, size, audience, gender, ageGroup - structured attributes where applicable
• additionalProperty - an array of PropertyValue objects for arbitrary attributes (waterproof rating, fit, use case, certifications). This is the field where ACO work lives for attribute coverage beyond the standard vocabulary.

AI engines reason about entities, not strings. Product schema makes each product an entity with structured attributes the engine can query directly. An AI answering "waterproof hiking boots under $200 for flat feet" can filter candidate products by offers.price, additionalProperty (waterproof = true), and audience (fit = wide) - the exact fields the schema exposes. Without schema, the engine has to guess at these attributes from free-text descriptions, which is slower and less reliable.

The measured lift is substantial:

• 2.5x more likely to be cited in AI Overviews with complete product schema (BrightEdge / Alhena.ai, 2025)
• 3.1x more frequent citation in Google AI Overviews for schema-compliant pages (Geolikeapro study, 2025)
• 61.7% citation rate for attribute-rich schema versus minimal markup (independent structured data research, 2025)

Schema is also how products enter the knowledge graphs that AI engines consult when composing shopping answers. Google's Knowledge Graph, in particular, is the substrate AI Overviews and AI Mode draw from for product answers. A product with rich schema becomes an entity in that graph; a product without schema is effectively invisible to the graph layer, even if its page ranks well in traditional search.

For retailers running ACO programs, product schema is the highest-leverage technical baseline. An ACO program on a catalog with no schema is optimizing a car with no engine - the downstream work only returns once the schema foundation is in place.

The recurring mistakes across ecommerce schema audits:

1. Mismatched data. The price in the schema is $99 but the rendered page shows $89. Google treats this as a spam signal and can apply a manual action. Schema must match the visible page, always.
2. Missing gtin / mpn. Without a global product identifier, your product cannot be confidently merged with the same product from other sources in the knowledge graph. This is one of the most common reasons a product page ranks well but does not appear as a product card in AI answers.
3. Stale priceValidUntil. priceValidUntil must be set to a future date. A past date tells Google the offer expired and the product drops from shopping results.
4. Thin description field. A 20-word description gives the retrieval layer almost nothing to match against. Invest in 150-300 word descriptions that cover material, fit, use case, and differentiators.
5. Variant inconsistency. Product pages with size/color variants must keep the JSON-LD schema, rendered HTML, and Google Merchant Center feed aligned on the selected variant. Variants that exist in the feed but not in the schema (or vice versa) cause the product to disappear or appear incorrectly.
6. Missing aggregateRating when you have reviews. If reviews exist on the page, aggregateRating must be populated. Missing it is a significant negative signal for AI product cards.
7. Not updating availability in real time. If schema says InStock and checkout fails, every engine that cached the schema gets a bad signal. For high-SKU retailers, dynamic schema rendering from the live inventory system is the only reliable approach.

Tools like Google's Rich Results Test and Schema.org Validator catch structural errors. They do not catch mismatch-between-schema-and-page errors, which is why a quarterly audit that spot-checks rendered page vs schema vs feed is a standard part of a healthy ACO program.