Product Schema
Product schema is structured data, almost always written as JSON-LD, that marks up a product's name, price, currency, availability and review data on a page using the Schema.org vocabulary. It gives crawlers a machine-readable version of facts that would otherwise have to be scraped from prose or a price widget, which is why it's the standard way to make a product page's own details legible to both search engines and AI systems.
Why it matters
Set expectations correctly here: Google's own guidance on optimizing for AI Overviews and AI Mode says no special markup, schema, or AI-specific file is required to be surfaced, and explicitly warns against writing content "for AI" instead of for people. So product schema is not a confirmed independent ranking or citation factor on its own. What it reliably does is remove ambiguity. A crawler reading unstructured HTML has to guess whether a number is a price or a SKU, and whether "in stock" is current or copy-pasted from a template. Product schema states these facts directly, which is also the mechanism behind classic rich snippets, price and star ratings shown in search results. For an AI assistant assembling an answer, correctly labeled facts are easier to lift cleanly than facts it has to infer, even though the schema itself doesn't buy you a place in the answer.
How it works
In practice, product schema is a JSON-LD block placed in a page's source, invisible to visitors, that declares an object of type Product with properties such as name, brand, offers (containing price, priceCurrency and availability) and aggregateRating. Schema.org is the shared vocabulary behind this, maintained jointly by Google, Microsoft and other engines, so the same markup is readable by more than one consumer. A crawler parses the block and takes the values as declared, rather than trying to extract them from visible copy. The one hard rule is that the marked-up values have to match what a visitor actually sees on the page; a JSON-LD block is not a place to state a price or availability status that differs from the live page.
Common mistakes
The most damaging mistake is markup that drifts from the visible page: an old price left in the JSON-LD, or a rating that isn't backed by real reviews. Search engines treat that as manipulation and can act on it. Almost as common is incomplete markup, no price or no availability property, which simply forfeits the rich-snippet treatment. Stale data causes quieter damage: a product sells out, the page updates, but the schema still declares InStock because nothing re-generates it automatically. Run every template through a validator, and where possible generate price and availability values from the same data source that renders the visible page, so the two can't fall out of sync. Duplicate or conflicting Product blocks on one page is the other common failure, and it tends to make crawlers trust the whole page's markup less.
Relation to AI recommendations
AI assistants draw on sources that are easy to extract facts from cleanly, and a well-formed product schema is exactly that: brand, price and rating stated as data rather than left to interpretation. That doesn't make it a ranking lever in the way backlinks or brand mentions are; an Ahrefs analysis of roughly 75,000 brands found web-mention frequency correlates with AI citation rate about three times more strongly than backlinks do, and structured data wasn't part of that comparison at all. Where product schema earns its keep is downstream of citation: once a page is being considered, clean facts reduce the odds an assistant misstates your price or invents an availability status, which matters given how often AI tools get sourced facts wrong. It's a hygiene factor for accuracy, not a visibility strategy on its own.
Example
A small bicycle shop in Rotterdam sells a trekking bike for 899 euros. Its product page carries a Product schema block with name, brand, price, currency, an availability value of InStock, and an aggregateRating built from its own verified reviews. If a shopper asks an AI assistant to suggest a solid trekking bike under 1,000 euros, the assistant can lift the price and rating from the markup instead of parsing them out of body copy, and state them correctly if the page is used at all. Without the schema, the assistant would have to infer the price from running text, and would be more likely to get it wrong or skip the shop's listing rather than risk it.
Common questions
Do I need to write JSON-LD by hand?
Usually not. Platforms like Shopify and WooCommerce generate Product schema automatically or through plugins. For a custom-built page, a JSON-LD generator will produce the block for you; you still need to keep its values wired to the same data that renders the visible price and stock status.
How do I check that my product schema is valid?
Run the page through Google's Rich Results Test or the Schema Markup Validator; both flag errors, warnings and missing required properties. Then check by hand that the marked-up price and availability actually match what's visible on the page, since a validator won't catch that kind of mismatch.