Structured data

Spencer Thursfield · Published 2026-09 · Last reviewed 2026-09-10 · Evidence: platform documentation

Structured data is a standardised, machine-readable description of a page, written in a shared vocabulary and placed in the page’s own code. It restates what a reader can already see, in a form a machine does not have to infer.

The modesty is the point. Structured data is the most oversold subject in AI search, and two of the claims made for it sit in our folklore ledger with the evidence against them.

What is structured data?

Google calls it “a standardized format for providing information about a page and classifying the page content”. The vocabulary is schema.org, a shared set of type names and properties. Pick the type that honestly describes the thing, fill in its properties, and put it on the page it describes.

Three syntaxes are supported, all equally acceptable to Google when valid. Google recommends JSON-LD, a block of JavaScript object notation in a <script> tag, as “the easiest solution for website owners to implement and maintain at scale”. Official recommendation

What does structured data actually do?

Three things are documented, each narrower than the way it is usually sold.

• It makes a page eligible for a richer appearance. Google’s catalogue of supported features runs to dozens of types: article, breadcrumb, event, product. Markup is the entry ticket, nothing more. Google states it plainly: “Using structured data enables a feature to be present, it does not guarantee that it will be present.” Official recommendation
• It states facts a machine would otherwise have to guess. Google uses markup “to understand the content of the page, as well as to gather information about the web and the world in general”. A trading name, an address, an author, an opening time: written once, unambiguously. Official recommendation
• It helps an entity resolve. The sameAs property is defined by schema.org as the “URL of a reference Web page that unambiguously indicates the item’s identity”, such as a Wikipedia or Wikidata entry. For a business sharing its name with three others it is the cheapest disambiguation there is; the argument is made in full in what an entity is. Strongly supported

None of it is a lever on position. Google’s clearest statement arrives in the guidelines on penalties: a structured data manual action “means that a page loses eligibility for appearance as a rich result; it doesn’t affect how the page ranks in Google web search”. If getting markup badly wrong does not move rank, getting it right is unlikely to either.

Does structured data help you appear in AI answers?

No. The field says otherwise loudly, and Google has answered it in writing.

You don’t need to create new machine readable files, AI text files, or markup to appear in these features. There’s also no special schema.org structured data that you need to add.

Google Search Central, AI features and your website

The same page sets the eligibility bar for AI Overviews and AI Mode: a page “must be indexed and eligible to be shown in Google Search with a snippet”, and “there are no additional technical requirements”. Markup is not part of that gate.

Structured data does appear in Google’s list of practices worth keeping up for AI features, in exactly one form: “making sure your structured data matches the visible text on the page”. It asks for accuracy, not addition.

The verdicts differ. There is no AI vocabulary, so a product selling one is selling a thing the platform says does not exist. Whether markup correlates with citation is a live question nobody has answered. We record schema presence and type on cited and uncited pages in our own benchmark, so we expect to answer it with numbers.

Which structured data should a small business have?

A short list, deliberately. Every block you add has to stay true.

• Organization, or LocalBusiness if customers visit you. Trading name, address, telephone, opening hours, and sameAs links to profiles you genuinely control. This is the identity block, the one that earns its keep. Official recommendation
• Person for anyone you name as an author. It connects a byline to a findable human, the mechanism described in author identity and provenance.
• WebSite once, defined by schema.org as “a set of related web pages and other items typically served from a single web domain”.
• The type that honestly describes the page. Article on an article, Event on an event, Product on a product. Google asks you to “use the most specific applicable type and property names”.
• FAQPage, with one caveat. Google removed its FAQ rich result documentation in June 2026, because “the FAQ rich result feature is no longer shown in Google Search results”. The markup no longer buys an appearance. Describing visible questions and answers is now the only reason to use it.

Nothing on that list requires a plugin, a subscription, or a consultant.

Why the markup must match the visible page

This is the rule that decides whether structured data helps or harms, and automated tools break it constantly. Google’s guidelines are direct: “Don’t mark up content that is not visible to readers of the page”, and “your structured data must be a true representation of the page content”. A rating nobody left, or an address you do not trade from, is not an optimisation. It is grounds for a manual action.

The test is one question: could a reader find every fact in your markup by reading the page? If not, either put it on the page or take it out of the markup. Required

How do you check it is valid?

• The Rich Results Test parses a URL or a block of code, reports errors and warnings, and names the features the page is eligible for. Run it before publishing.
• URL Inspection in Search Console shows what Googlebot actually received, which is the only version that counts, and matters most where markup is injected by a script.
• The rich result report in Search Console tracks valid and invalid items across the whole site over time, which is how you catch a template change that silently broke a hundred pages.

For whether it was worth doing at all, Google advises a before-and-after test: take stable pages with several months of Search Console history, add markup, confirm Google has found it, then read the performance report per URL.

What we do not know

Whether the presence of structured data correlates with selection as a source in a generated answer. It is measurable, we are measuring it, and nobody publishing confident advice on it appears to have done so. Whether markup carries weight in systems that document nothing, such as ChatGPT search or Perplexity, is also unknown, and their silence is evidence in neither direction.

Common questions

Is structured data a ranking factor?

Google does not describe it as one. Its documentation says something narrower: markup makes a page eligible for a rich result, and a structured data manual action removes that eligibility without affecting how the page ranks.

Is there special schema for AI Overviews, AI Mode or ChatGPT?

No. Google states there is no special schema.org structured data you need to add for its AI features, and no other platform publishes such a vocabulary either. A product sold as AI schema is selling something that does not exist.

Does adding structured data make an AI system cite you?

No platform claims it and no adequate public evidence shows it. Our benchmark records schema presence on cited and uncited pages, and we will publish what it shows.

Do I still need FAQPage markup?

Google removed the FAQ rich result documentation in June 2026 because the feature is no longer shown. The markup no longer buys an appearance. Describing questions a reader can see on the page is the only remaining reason to use it.

What is the minimum a small business should add?

One Organization or LocalBusiness block with your name, address, contact details, opening hours and sameAs links to profiles you control; Person for anyone you name as an author; and whichever type honestly describes each page.