Product schema: definition and examples

In detail

Product schema is defined at schema.org/Product and covers the properties machines use to resolve a product's identity, price, and quality signal. The common implementation is JSON-LD — a script block containing a Product object with name, brand, sku, gtin, description, image, offers, and optionally aggregateRating. The same data can be expressed in microdata or RDFa; JSON-LD is the convention because it is cleanest to maintain and the format Google's product structured data docs recommend.

Not every field is equal. For AI visibility, offers (price, priceCurrency, availability), sku, gtin, and aggregateRating matter most; name, description, and image are table stakes. Fields like material, color, and size become decisive on faceted queries — "best ceramic mugs that are dishwasher safe" depends on the engine reading material and care data without inference.

For agent-led commerce, the same block does double duty: machine-readable price and availability are core inputs to the agent-readability score, and protocol-based checkout (ACP, UCP) assumes the catalog data agents read upstream is accurate.

Why it matters for ecommerce

Product schema is how an AI engine reads a product without guessing. A PDP with thorough schema hands the engine answers to half the sub-queries in a query fan-out before any free-text parsing happens. A PDP with sparse schema forces inference — which is where misses and hallucinated specs enter the answer.

For a Shopify brand it is also the lowest-effort, highest-yield PDP improvement available: one theme or metafield-backed app change that populates three missing fields catalog-wide can move visibility metrics within weeks, because it repairs every PDP at once instead of one page at a time.

Example

A ceramic-mug brand's 12oz hand-thrown mug emits Product JSON-LD with name, description, image, and an offers block — but no gtin, material, aggregateRating, or populated category. Adding the four fields through a metafield-backed schema app lifts the SKU's ChatGPT share of model for "best ceramic mug for a gift" prompts from 6% to 18% over the next two weekly runs (illustrative, but the mechanism — review signal plus resolvable identity — is the documented one).

How eCommerce Insights checks it

Product JSON-LD completeness is the largest input to the agent-readability score: every SKU is parsed for the AI-relevant fields, missing or malformed properties are flagged, and the fix ships as a reviewable diff — often a metafield mapping rather than a code change. The free Product Schema Generator builds a correct block from a product URL, and PDP optimization covers the full workflow.

Related terms

PDP optimization — the broader practice schema sits inside.
SKU-level AEO — the discipline schema most directly supports.
Agent-readability score — where schema completeness is scored.
Query fan-out — why field coverage wins sub-queries.
llms.txt — the unstructured counterpart for LLM crawlers.

Ask AI about product schema

Have your favorite AI engine apply this definition to your catalog.

Frequently asked questions

Which Product schema fields actually matter for AI search?

name, brand, sku, gtin where applicable, description, image, offers (with price, priceCurrency, and availability), and aggregateRating when real review data exists. category and material matter for faceted retrieval. aggregateRating is the field that most often separates cited SKUs from uncited ones, because AI engines lean on review evidence when choosing which products to surface.

Do Shopify themes emit Product schema correctly out of the box?

Partially. Most themes emit a minimal Product block — name, image, price — but skip gtin, material, category, and review aggregation, and some emit conflicting blocks when a reviews app injects its own. The practical move is to audit what the rendered page actually emits rather than trusting the theme documentation.

Is JSON-LD required, or does microdata work?

Schema.org vocabulary can be expressed as JSON-LD, microdata, or RDFa, and engines read all three. JSON-LD is the convention because it lives in one script block, is easiest to maintain programmatically, and is the format Google's documentation recommends. There is no AI-visibility reason to choose anything else.

Can Product schema fix a thin PDP by itself?

No. Schema makes existing product data machine-readable; it does not create answer coverage. A PDP with complete JSON-LD but a two-sentence description still loses sub-queries about fit, care, and use cases. Schema plus answer-shaped copy is the combination that wins citations.

How do I check what schema my PDP currently emits?

Run the rendered URL through a structured-data validator, or use a tool that audits the AI-relevant fields specifically. The free eCommerce Insights AEO Grader reads any public PDP and reports which Product fields are present, missing, or malformed, weighted by how much each matters for citation.

Go deeper

Generate a complete Product JSON-LD block from any product URL with the free Product Schema Generator.