Schema Markup (Structured Data)

What does schema markup still genuinely earn?

Rich results, entity plumbing and one first-party claim. The rich results are real, documented and worth having, and Google’s gallery is the authoritative list of which types produce them. Organization and Person markup with sameAs is how a string becomes a resolvable entity. And Microsoft has stated that schema helps its language-model pipeline understand content, which is the only first-party statement of that kind from any AI platform.

Worth knowing before you build a checklist from an older guide: neither FAQPage nor HowTo appears in that gallery any more, and those two are the types most AI-era advice recommends first.

Why does schema not earn AI citations?

Because it has been tested and it does not. Ahrefs added schema to 1,885 pages against roughly 4,000 controls over seven months and measured AI Overviews down 4.6%, AI Mode up 2.4% and ChatGPT up 2.2%, which is indistinguishable from zero. A separate test found no engine extracted a fact that existed only in JSON-LD, which is consistent with training pipelines stripping script tags before anything is embedded.

Google’s own guidance says the same thing from the other direction: no special structured data is needed for its AI features. The industry sold the opposite on the strength of a correlation, and the correlation was site quality: large well-run sites have schema and citations, and adding schema to a small site adds one of the two.

How should you emit it, given all that?

From the same fields that render the visible copy, and never as a place for a fact to live. If a price, a date or a specification has to be quotable, it belongs in the body text first and in the markup as a mirror. Schema describing content a page does not display is a manual-action risk and is invisible to every assistant at the same time.

The recommendation: keep emitting it, generate it, and stop reporting it as an AI visibility action. It costs nothing when it comes from the same source as the page, and its budget line should sit under rich results rather than under AI. One exception worth making: Organization and Person markup with real sameAs links is entity work rather than rich-result work, and it earns its place on a different argument entirely. And one implementation rule that prevents the whole category of drift: generate the markup from the same values that render the visible page. Two hand-written copies of a fact eventually disagree, and the one nobody can see is the one that stays wrong.

Related terms

• Entity (SEO)An entity is a distinct thing a search system has identified and can reason about: a company, a person, a product, a place.
• Knowledge GraphThe Knowledge Graph is Google’s database of entities and the relationships between them: people, companies, places, works and the facts connecting them.
• Answer Engine Optimization (AEO)Answer Engine Optimization is the practice of writing pages so an AI assistant can lift a direct answer out of them.
• E-E-A-TE-E-A-T stands for experience, expertise, authoritativeness and trustworthiness, a framework from Google’s search quality rater guidelines.