Structured data: definition and its role in AI search

99Visibility Research · Updated October 21, 2026 · 4 min read

Short answer

Structured data is machine-readable markup that describes what a page contains, usually written as JSON-LD using the schema.org vocabulary. Search engines parse it to understand entities such as products, prices and organizations. Google says structured data isn't required for generative AI search. Existing markup does no harm, but no published evidence shows it is a lever for getting cited in AI answers.

Key takeaways

Structured data is a labelled description of page content, written for parsers, not for readers.
Google states structured data isn't required to appear in AI Overviews or AI Mode, and no AI-specific schema type exists.
Ahrefs' matched test of 1,885 pages found no major citation uplift after adding schema, and AI Overviews citations fell 4.6%.
Keep valid markup you already have, but don't expect it to put you in AI answers.
The facts that matter should sit in server-rendered page text, where every crawler can read them.

Definition

Structured data is code on a page that states, in a fixed vocabulary, what the page is about. A product page might declare its name, price and availability. An about page might declare the organization's name, logo and address.

Most sites write it as JSON-LD, a block of JSON in the page head, using types from schema.org. Microdata and RDFa are older formats that do the same job inside the HTML. The visible page and the markup should say the same thing.

This entry is a short definition. For the vocabulary itself, see schema markup. For the full evidence review, read does schema markup help AI citations.

How it works

A crawler fetches the page, finds the markup and parses it into fields. Google's documentation describes using it to understand page content and to power certain rich results in classic search.

Three practical points follow:

Eligibility is not display. Valid markup makes a page eligible for a rich result. It doesn't guarantee one.
Features get retired. Google deprecated FAQ rich results in May 2026 and removed the FAQPage documentation in June 2026.
Rendering matters. Vercel found that ChatGPT and Claude crawlers fetch JavaScript files but don't execute them. Markup injected by client-side scripts may never reach those crawlers. Googlebot renders JavaScript, so this gap is mainly about non-Google AI crawlers.

If you want to confirm your markup matches what the page says, a JSON-LD checker will show mismatches.

Why it matters for AI visibility

As of October 2026, Google's position is direct. Its AI features page says a page needs to be indexed and eligible for a snippet to appear in AI Overviews or AI Mode, with no additional technical requirements. Its generative AI optimization guide says you don't need new markup to appear, because Google Search doesn't use special files for this.

The best test available agrees. Ahrefs tracked 1,885 pages that added schema against matched controls and found no major uplift on any platform. AI Overviews citations fell 4.6%, a statistically significant drop. Our digest of the Ahrefs study covers the method.

One boundary matters. Every page in that study already had 100 or more AI Overview citations. Whether schema helps a page that isn't cited at all is untested, not disproven.

What this means: treat structured data as classic search hygiene. Spend AI visibility effort on access first, then on stating your key facts in plain, fetchable text. The trade-off is laid out in structured data vs plain-text facts.

Common misconceptions

"Add schema to get cited by AI." Google says it isn't required for generative AI search, and Ahrefs' matched test found no lift.
"FAQ schema wins AI answers." FAQ rich results are deprecated. Question-and-answer content on the page is a separate matter from the markup.
"Schema is useless, so remove it." Google says existing markup does no harm. Removing it is not supported by evidence either. See whether removing schema hurts AI visibility.
"Markup replaces page text." A crawler that can't see your price in the HTML gets no quotable sentence from it. State the fact in the text.

Frequently asked questions

Is JSON-LD the same as structured data?

JSON-LD is one format for writing structured data. Others include Microdata and RDFa. Google supports all three for classic search features, and JSON-LD is the most common because it sits in one block instead of being woven through the HTML. The format choice has no documented effect on AI answers. What matters more is that the markup is accurate and present in the HTML that crawlers actually receive.

Should I add structured data to a new site?

Add it if you want classic rich results Google still supports, such as product details. Don't add it expecting AI citations, because Google says it isn't required for generative AI search. For a new site, the higher-value checks are whether AI crawlers can reach your pages, whether your key content renders without JavaScript, and whether your pages are indexed and eligible for snippets.

Does ChatGPT or Perplexity read structured data?

Neither OpenAI nor Perplexity documents using schema markup to select or cite pages. Ahrefs measured a 2.2% change for ChatGPT citations after pages added schema, within an overall finding of no major uplift on any platform. Since many AI crawlers don't execute JavaScript, markup added by client-side scripts may not be seen by them at all.

Can wrong structured data cause AI to state wrong facts?

No study in our evidence base measures that directly. Contradictions inside your own site, such as one price in the markup and another on the page, are still worth fixing. They create conflicting signals about the same fact, and fixing them is fully in your control. Keep markup, visible text and every page that states the fact in agreement.

Sources

This page is AI-assisted. It was written from a verified evidence base and checked against our editorial policy. Engine behaviour changes often; facts are as of October 21, 2026.