Blog · Answer Checkup team · Last checked

Does schema markup help AI search? What the evidence says

A lot of AI visibility advice starts with "add schema." We check for it in our own free scan, so we wanted an honest answer to a simple question: does schema.org markup get you cited in AI answers?

Short version: there's no good evidence that adding it, on its own, gets you cited more. There are still good reasons to have it. Here's what we found, and what we'd do.

What schema markup is

Schema markup is structured data, usually a JSON-LD block in your page's HTML, that labels things in a format machines can parse: this is an Organization, this is its logo, this is a Product with this price. Google recommends JSON-LD among the three formats it supports.

For years its main job was rich results in Google: stars, prices, event dates. The new question is whether AI answer engines use it to decide what to cite.

What Google says

Google's page on AI features is direct. To appear in AI Overviews or AI Mode, "There are no additional requirements" beyond being indexed and eligible to show with a snippet. It also says: "There's also no special schema.org structured data that you need to add."

The same page lists ordinary SEO basics that still matter, including "Making sure your structured data matches the visible text on the page."

So Google's position is: schema isn't a ticket into AI features, but if you use it, it must be accurate.

What the tests found

A large before-and-after study. Ahrefs tracked 1,885 pages that added JSON-LD between August 2025 and March 2026 and matched them against 4,000 control pages. They measured citations in Google AI Overviews, AI Mode and ChatGPT. Their finding: "Adding schema produced no major uplift in citations on any platform."

The same study explains why the myth is so sticky. In a first pass over 6 million URLs, they found AI-cited pages were "almost three times more likely to have JSON-LD than non-cited pages." But sites that add schema also tend to do the rest of SEO well. That's correlation, not cause.

One limit worth knowing: every page in that study was already being cited a lot before it added schema. Ahrefs says schema "might still play a role" for pages AI systems aren't seeing at all. Their study can't answer that.

Live retrieval tests. Ahrefs also points to an experiment by searchVIU that checked whether five AI systems used schema when fetching a page in real time. The result as Ahrefs reports it: "every system extracted only visible HTML content." JSON-LD and hidden markup were ignored.

Search Engine Roundtable wrote up two LinkedIn posts on the same theme from September 2025. In one test, two pages described the same made-up product, one with visible text plus structured data, one with structured data only. According to the tester, the models could pull details like price and colors only from the page where the information was visible as text.

Where schema is said to help. Search Engine Land reports that Fabrice Canel of Microsoft Bing confirmed in March 2025 that schema markup helps Microsoft's LLMs understand content for Copilot. The same article is careful to say other AI platforms haven't confirmed how, or whether, they use it.

What we take from it

  1. Visible text does the work. If your price, service area or opening hours only live in JSON-LD, assume an assistant fetching your page won't see them. Put them in the page.
  2. Schema is labeling, not a lever. It can make your facts easier to parse for systems that do read it, like Google's rich results. It won't make a weak page get cited.
  3. Wrong schema is worse than none. Google's guidelines say not to mark up content that isn't visible to readers, and that structured data must be a true representation of the page.

What we'd still do, in order

This takes an afternoon on most small sites.

  1. Write the facts in plain HTML first. Who you are, what you sell, where, and how to reach you. Make sure they're in the raw HTML, not only added by JavaScript. Our post on JavaScript and AI crawlers shows how to check.
  2. Add one Organization or LocalBusiness block to your homepage. Name, URL, logo, contact details, and sameAs links to profiles you actually own. We cover this in entity signals.
  3. Add Product or Service markup where you list what you sell, and Article on blog posts. Use the most specific type that fits. Google asks for that too.
  4. Put it in the HTML, not only in a script that runs later. Ahrefs notes that AI crawlers appear to treat JSON-LD in the page HTML and JSON-LD injected by JavaScript differently.
  5. Test it. Google's Rich Results Test and URL Inspection tool catch most technical errors.
  6. Then stop. Spend the rest of the time on the page content and on earning real mentions on other sites. That's where the evidence points.

Why our scan still checks it

Fair question, given all of the above. Our free scan scores six checks out of 100 points:

  • AI crawler access in robots.txt: 30
  • Main content in the raw HTML: 20
  • Structured data: 15
  • Title and meta description: 15
  • llms.txt: 10
  • Sitemap: 10

Structured data splits 8 + 7. The 8 is for at least one JSON-LD block that parses. The 7 is for a key type such as Organization, Product, SoftwareApplication, LocalBusiness or WebSite. The biggest weights go to the things that block crawlers outright: robots.txt rules and content missing from the raw HTML.

We keep it because it's cheap, Google uses it for rich results, Bing has said it helps Copilot, and broken or missing markup is often a sign of other technical gaps. We don't claim it moves AI citations. A perfect scan score doesn't mean AI answers will mention you.

If you want to know whether answers mention you at all, that's what our paid reports are for. Our free scan reads your site itself. It does not ask any AI engine. Our paid reports, coming soon, ask buyer-style questions through the OpenAI API with web search and the Perplexity API, and show your share of voice and the domains those answers cite. API answers are close to what people see in the consumer apps, but they are not the same thing, so we treat them as a sample, not a mirror.

FAQ

Does ChatGPT read schema markup?
OpenAI hasn't published how its search uses schema. In the live retrieval test Ahrefs cites, the systems tested used only the visible HTML.
Does schema help with Google AI Overviews?
Google says no special schema is needed. Structured data must match the visible text if you use it.
Should I add FAQPage schema for AI?
Only if the page really has those questions and answers in visible text. Treat it as tidy labeling, not a ranking trick.
Is schema bad for AI visibility?
No. The evidence says it mostly doesn't move AI citations either way. It still helps with accuracy and with Google's rich results.

Sources

All fetched Oct 15, 2026.

Check your own site

Curious whether your structured data, robots.txt and raw HTML are readable by AI crawlers? Run the free scan.

Answer Checkup is not affiliated with OpenAI, Perplexity, Google or Microsoft.