Does structured data help you get cited in AI answers?
No measurable lift so far. Google says AI features need no special schema, and an Ahrefs test of 1,885 pages that added JSON-LD found no clear change in AI citations.
The short answer
There is no good evidence that adding structured data gets a page cited more often in AI answers. Google says its AI features need no special schema.org markup, and the one controlled test published so far, Ahrefs’ study of 1,885 pages that added JSON-LD, found no clear positive effect in AI Overviews, AI Mode or ChatGPT. Keep markup for what it is documented to do (rich results and helping search engines read your page) and spend the AI effort on the visible text.
Why do so many people think schema helps AI answers?
Because cited pages usually have it. Two large correlation studies point the same way.
Ahrefs looked at 6 million URLs and found that pages cited by AI were almost three times more likely to carry JSON-LD than pages that were not cited. Semrush, in a study published on 5 January 2026, analysed 5 million URLs cited by ChatGPT Search and Google AI Mode and reported how often each markup type appeared on them:
| Markup type | Share of cited pages, ChatGPT | Share of cited pages, AI Mode |
|---|---|---|
| Organization | 25% | 34% |
| Article | 20% | 26% |
| BreadcrumbList | 15% | 20% |
| FAQ | 3% | 5.5% |
| Product | 1.5% | 2.5% |
Semrush says plainly that “this study identifies correlations, not causation.” Ahrefs gives the likely reason for the pattern: schema “tends to live on better-maintained, more technically sophisticated sites.” Those sites also rank well, publish more and get crawled more, and any of that could explain the citations.
Microsoft is the other source of the belief. In March 2025 Search Engine Roundtable reported that Bing’s Fabrice Canel told an SMX Munich audience that schema markup helps Microsoft’s LLMs understand content. The report came from David Mihm’s LinkedIn summary of the talk; Canel replied to the post without disputing it. There is no transcript or Microsoft documentation behind it, and “helps understand” is a weaker claim than “earns citations.”
What does Google say about structured data and AI Overviews?
Google has answered this directly, twice.
Its guide to generative AI features on Google Search says: “Structured data isn’t required for generative AI search, and there’s no special schema.org markup you need to add.” The same guide says you don’t need to create “machine readable files, AI text files, markup, or Markdown” to appear in Google Search, including its generative AI features.
Its page on AI features and your website sets the 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.” Structured data appears there only as an ordinary best practice: make sure it matches the visible text on the page.
Google does not tell you to remove markup. The AI guide says to keep using it “as part of your overall SEO strategy” because it helps with rich result eligibility. And Google’s general structured data documentation says it uses markup “to understand the content of the page.” So the markup is read; Google just does not describe it as an input to being chosen for an AI answer.
What happened when pages added schema?
Ahrefs published the closest thing to a controlled test on 11 May 2026. The design:
- Take 1,885 pages that added JSON-LD between August 2025 and March 2026, confirmed by checking each page’s HTML history for the date schema first appeared.
- Match each one with three control pages from other domains that never added JSON-LD and had similar citation levels beforehand, about 4,000 controls in total.
- Count citations in Google AI Overviews, AI Mode and ChatGPT for 30 days before and 30 days after the change, and compare the two groups with a difference-in-differences model.
The results, relative to the controls:
| Platform | Change in citations | What Ahrefs concluded |
|---|---|---|
| Google AI Mode | +2.4% | Indistinguishable from zero |
| ChatGPT | +2.2% | Indistinguishable from zero |
| Google AI Overviews | -4.6% | Statistically significant, cause unclear |
Ahrefs’ summary: “schema had no clear positive or negative effect.” It does not claim schema hurts AI Overview citations. Both groups were already declining before the change, outliers drove much of the gap in one of its robustness tests, and the absolute loss was about 12 daily citations per page.
The authors also cite a separate test by searchVIU in which five systems (ChatGPT, Claude, Perplexity, Gemini and Google AI Mode) fetched a page live and none of them used its schema markup; each read only the visible HTML. We could not open searchVIU’s original write-up, so that result is reported here as Ahrefs describes it.
What are the limits of that evidence?
They are real, and Ahrefs lists most of them itself.
- Every page in the sample already had more than 100 AI Overview citations before the markup went in. The study says nothing about pages that AI systems do not yet surface, and Ahrefs says schema “might still play a role” there.
- Schema types were pooled. Article, FAQ, Product, HowTo and Organization markup were measured together, so a type that helps could be hidden by types that do not.
- Thirty days may be too short for slow effects such as a model learning about an entity over time.
- Only JSON-LD in the HTML was tested, not Microdata, RDFa or markup injected by JavaScript.
- Sites often add schema at the same time as other fixes, so the treatment is never perfectly isolated.
So the honest reading is narrow. For pages that are already cited, adding JSON-LD did not move citations in a month. Nobody has published a test showing that it does.
What does structured data still do?
Several documented things, none of them an AI citation boost.
| Use | Who documents it | Worth doing? |
|---|---|---|
| Rich result eligibility in Google Search | Google Search Central | Yes, for types Google still shows |
| Helping Google understand the page and the world | Google Search Central | Yes, as part of normal SEO |
| Merchant and product details | Google (Merchant listing markup, Merchant Center) | Yes for ecommerce |
| A citation boost in AI answers | No platform documents this | Do not count on it |
The rich result list keeps shrinking, which matters when you decide what to maintain. Google’s documentation changelog shows it removed the practice problem type in January 2026 and FAQ rich results in 2026: a deprecation notice on 8 May said the feature would stop appearing from 7 May, and the documentation was removed on 15 June. Several other types, including course info, estimated salary and vehicle listing, went in September 2025. Markup for a retired feature does no harm, but it earns nothing in Google Search either.
Commerce is the case where structured information clearly matters to AI answers. Google’s AI guide says generative responses can include product listings and local business information, and points site owners to Merchant Center feeds and Google Business Profiles for that. The guide names feeds and profiles here, not schema on the page.
What should you put on the page instead?
Text a reader and a retriever can both use. The evidence above points to one shared pattern: the systems that build AI answers read the visible page.
Google’s own list is short. Allow crawling in robots.txt and at your CDN, keep important content “available in textual form”, support it with good images and video, and keep structured data consistent with the visible text. Google also says there’s “no requirement to break your content into tiny pieces” for AI.
If a fact only lives in your JSON-LD, such as a price, a date or an author’s role, assume an AI assistant fetching the page will not see it. Put it in the text as well.
What to do
- Keep the markup you have if it is valid and matches the page. Do not rip it out to chase the AI Overviews result in the Ahrefs table.
- Add new types only where Google still shows a rich result or where you sell products, and check the documentation changelog before investing in a type.
- Move every fact that matters (prices, dates, specifications, authorship) into the visible text, not just the JSON-LD.
- Confirm your key pages are indexed and snippet-eligible in Search Console, since that is Google’s only stated requirement for AI features.
- If you want to know whether schema helps your own pages, test it: add markup to a set of pages, keep a matched set unchanged and compare citations after 30 days, as Ahrefs suggests.