Does structured data get you cited by AI?
Schema markup is the most repeated piece of AI-visibility advice, often sold as the single biggest change you can make. When it has been tested properly, it has not held up. Here is what the evidence shows, what structured data is still good for, and where the effort belongs instead.
What a controlled test found
Most claims about schema and AI come from comparing sites that happen to have it with sites that don't, which mostly measures how well-run those sites are. Ahrefs ran a matched comparison instead: 1,885 pages that added structured data against about 4,000 similar pages that did not, tracked over the same period. Pages that added schema saw AI Overviews citations fall by 4.6%. Citations in AI Mode rose 2.4% and in ChatGPT 2.2%, small enough to be noise.
Google says the same thing plainly in its own documentation: there is no special schema.org markup you need to add for its AI features. One honest caveat: the pages in that test were already being cited. It does not prove schema is useless for a site no engine has found yet. But there is no good evidence that it is the way in either.
Sources: Ahrefs, matched study of schema and AI citations; Google Search Central guidance on AI features.
Why it doesn't move AI citations much
Being cited is mostly a question of being retrieved. When someone asks ChatGPT or Perplexity a question, the engine runs its own search and only quotes pages that search returned. Large studies of AI answers find that which pages get retrieved explains far more of who is cited than anything about the page itself. Markup describing a page does little to change whether that page is retrieved in the first place.
What structured data is still good for
- Rich results in Google. Review stars, prices, availability and breadcrumbs in ordinary search results, which can lift clicks.
- Unambiguous facts. A clean Organization block with your name, logo and profiles does no harm, and helps where another company shares your name.
- Products and local businesses. Product and LocalBusiness markup feed shopping and map surfaces that have their own requirements.
So keep it, keep it valid, and describe only what is really on the page. Just don't treat it as an AI strategy.
Where the effort belongs instead
- Let the right crawlers in. Citation runs through each engine's search crawler: Googlebot, Bingbot, OAI-SearchBot, PerplexityBot. Blocking training crawlers like GPTBot does not stop you being cited.
- Check your robots tags. Bing says NOARCHIVE keeps your content out of Copilot answers, and Google only uses snippet-eligible pages in AI Overviews, so nosnippet takes you out.
- Put your content in the HTML. The crawlers behind ChatGPT, Claude and Perplexity don't run JavaScript, so a page built in the browser looks empty to them.
- Answer the questions people ask in self-contained paragraphs under clear headings, on pages that can be found for those questions.
The free Avantas audit checks those gates directly and gives structured data almost no weight, for the reasons above. For the wider picture, see what AEO means and the full seven-lever guide.
Common questions
Does structured data help you get cited in AI answers?
Not measurably, on the best evidence available. In a matched test of 1,885 pages that added schema against about 4,000 that did not, Ahrefs found AI Overviews citations fell 4.6%, while AI Mode rose 2.4% and ChatGPT 2.2%, which is within noise. Google's own guidance says there is no special schema you need to add to appear in AI features.
So should I remove my schema?
No. Structured data still earns rich results in Google, such as review stars, prices and breadcrumbs, and those can lift clicks from ordinary search. Keep it for that. Just don't expect it to get you named in an AI answer, and don't spend effort on it ahead of the things that do gate AI visibility.
What should I fix instead?
The documented gates. Make sure the crawlers behind each engine are allowed in robots.txt, that pages are not set to noindex, noarchive or nosnippet, and that your content is in the HTML rather than built with JavaScript, because the crawlers behind ChatGPT, Claude and Perplexity don't run it.