Structured Data for AI Search: What the Evidence Actually Says

Someone has probably told you that without schema, ChatGPT and Google’s AI Overviews are just guessing who you are. Add the right code, the pitch goes, and AI tools will finally understand your business and start recommending it.

It’s a tidy story, and structured data for AI search is now sold on the back of it. The evidence is much less tidy. Google, which runs AI Overviews, says in writing that there’s no special schema you need to add. And the largest controlled test so far found that adding schema didn’t move AI citations.

First, the term. Structured data, usually called schema, is code in your page that labels facts for machines: this is our name, this is our phone number. It’s normally written in a format called JSON-LD.

This guide covers what Google says, what the tests found for ChatGPT and others, where the big percentages come from, and which schema a local business should still keep (and keep true).

What Google Says About Schema and AI Overviews

Google has answered this in writing, more than once.

Its AI features page (last updated December 2025) says: “There are no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary.” And: “There’s also no special schema.org structured data that you need to add.” A page just needs to be indexed and eligible to show with a snippet.

Its guide to optimising for generative AI, added in May 2026, repeats this under mythbusting: “Structured data isn’t required for generative AI search, and there’s no special schema.org markup you need to add. However, it’s a good idea to continue using it as part of your overall SEO strategy, as it helps with being eligible for rich results on Google Search.” The same section says Google Search doesn’t use llms.txt files.

A May 2025 Google post by John Mueller calls structured data something “our systems consider”, but says to make sure “all the content in your markup is also visible on your web page”. Mueller was blunter in April 2025: “Structured data won’t make your site rank better.”

To be fair, Google does use structured data “to understand the content of the page, as well as to gather information about the web and the world in general”. Schema isn’t useless; it just isn’t a special AI switch. For schema markup and AI Overviews, it’s ordinary SEO hygiene, not a separate requirement. Our guide to appearing in Google’s AI Overviews covers what does count.

Does Schema Help ChatGPT? What the Tests Found

“Does schema help ChatGPT?” sounds like opinion. It’s testable, and a few people have tested it.

searchVIU: what chatbots read on a page

In October 2025, searchVIU put fictional product prices in different places on a test page (visible text, JSON-LD only, microdata) and asked ChatGPT, Claude, Perplexity, Gemini and Google AI Mode for them. None found the price that existed only in JSON-LD. Its conclusion, published December 2025: “JSON-LD Schema is NOT read by AI chatbots during direct fetch.”

The limit: that’s direct page fetches only. searchVIU says schema may still be used when search engines index pages, and the test says nothing about training data.

Ahrefs: 1,885 pages that added schema

Ahrefs (May 2026) tracked 1,885 pages that added JSON-LD, each matched with three similar pages that hadn’t, comparing AI citations 30 days before and after. Against the controls, AI Overviews citations fell 4.6% (small, but statistically significant). AI Mode rose 2.4% and ChatGPT 2.2%, both statistically indistinguishable from zero.

Ahrefs’ own caveats: every page was already heavily cited by AI, so pages AI doesn’t know yet may differ. All schema types were pooled (“It’s possible some types help more than others”), the window was 30 days, only JSON-LD in the page HTML was studied, and they “can’t fully separate schema from” other changes made at the same time.

Fischman: a preprint, not a verdict

Kurt Fischman’s February 2026 paper is a single-author preprint: not peer reviewed, and correlational. Across 730 AI citations from ChatGPT and Gemini, schema presence had no significant effect; Google organic rank position was the strongest predictor. The fair exception: pages with Product or Review schema filled with concrete details (prices, ratings) were cited more. But a local service business can’t get review stars for reviews about itself anyway (more below).

How ChatGPT finds local businesses

OpenAI says ChatGPT search “sometimes partners with other search providers”, sending them rewritten versions of your question, and links to Microsoft’s privacy statement. For local answers it may use third-party providers of things like nearby business listings. Whether any of them use your schema isn’t published.

The counterweight: Microsoft Bing’s Krishna Madhavan wrote in October 2025 that “Schema is a type of code that helps search engines and AI systems understand your content.”

So no test shows schema hurts understanding. None shows it earns AI citations on its own.

Where the ‘Schema Gets You More AI Citations’ Numbers Come From

You’ve probably seen a precise-sounding figure for how much more often schema gets you cited by AI. Follow it back before you act on it.

One of the most repeated figures leads to an agency blog post naming no study, sample or method. Other articles repeat it, crediting that blog, and it shifts as it travels: all schema on one site, FAQ schema only on another, product pages on a third.

In August 2026, reviewer Jordan Kwan read 29 articles on schema and AI citations. Eight gave a specific number for schema’s effect; none traced back to a study that measured it. His verdict on the most repeated one: “An agency blog, citing an agency blog, citing nobody.”

Several pro-schema pieces are also written by schema vendors. That doesn’t make them wrong, but it’s worth knowing. Three questions sort evidence from marketing:

• Who ran the test, and when?
• Was there a comparison group that didn’t add schema?
• Can you read the method?

If the answer to any is no, treat the number as marketing. The tests above pass that check. The percentages in sales pitches don’t.

The Schema That’s Still Worth Having for a Local Business

None of this means ripping your schema out. Google’s own AI guide says keep using it, and two reasons hold up:

• Rich results. Schema makes a page eligible for Google’s enhanced results. Eligible, not guaranteed.
• Entity clarity. Google uses structured data to understand the page and “the world in general”. Organization or LocalBusiness markup states your name, address, phone and profiles in a form Google can read. Our guide to entity SEO for local businesses covers Organization, LocalBusiness and sameAs in detail.

So is LocalBusiness schema worth having for AI? Yes, but its job is describing you accurately to Google. It isn’t a lever for AI answers.

Existing FAQPage markup can stay. Google said in 2023: “Structured data that’s not being used does not cause problems for Search, but also has no visible effects in Google Search.” Just don’t pay anyone to add more. The supported list keeps shrinking, too: in June 2025 Google phased out seven more types, including Claim Review and Vehicle Listing.

For a trade or service business, accurate LocalBusiness and Organization markup covers it. Skip anything sold on the promise of stars or AI citations.

Keep It True: Schema That Contradicts Your Page Is a Liability

Say a plugin set up your schema years ago. It still says you’re open till 8pm on Saturdays and lists three towns you stopped covering.

That’s the real risk: wrong schema, not missing schema. Google’s rules say: “Don’t mark up content that is not visible to readers of the page.” Markup “must be a true representation of the page content”, and mustn’t be used “to deceive or mislead users”.

Break them and you can get a manual action, which means the page “loses eligibility for appearance as a rich result; it doesn’t affect how the page ranks in Google web search.” Google also clarified in May 2026 that its spam policies apply to generative AI responses in Search.

And remember searchVIU: a fact that lived only in schema wasn’t seen by chatbots reading the page. Put it on the page first.

A quick check:

1. Open your page source and search for application/ld+json.
2. Compare the name, address, phone, hours and service areas with the visible page and your Business Profile.
3. Remove anything untrue or not on the page, including your own review stars.
4. Run the page through Google’s Rich Results Test.
5. Check Search Console’s rich result reports after any changes.

It’s the business-facts step of a GEO audit. If nobody has opened your schema since a plugin wrote it, start there.

If AI Visibility Is the Goal, Put the Effort Here Instead

If schema isn’t the switch, what is? For a local business, the sources above point three ways:

• Your Google Business Profile. Google’s AI guide says Business Profiles “can help your products and services to be visible in both AI responses and other Google Search results”.
• Visible, specific facts on your pages. Services, towns covered, prices where you can give them. That’s what the tested AI tools actually read.
• Ranking in ordinary search. Fischman’s preprint found Google rank position the strongest predictor of being cited, and Google says its AI features rest on SEO fundamentals.

For the wider picture, see how AI search tools choose what to cite.

If you’d like someone to check what your schema says and whether it matches your site and Business Profile, tell us what you do and where. You’ll get a straight answer, and if we can’t help, we’ll say so.

FAQ

Do I need schema markup to appear in AI Overviews?

No. Google says there are no additional requirements to appear in AI Overviews or AI Mode, and no special schema.org structured data you need to add. A page needs to be indexed and eligible to show in Google Search with a snippet. Keep your schema for rich results, and keep it matching what’s visible on the page.

Does ChatGPT use schema markup?

There’s no evidence it reads JSON-LD when it fetches a page. In searchVIU’s October 2025 test, no chatbot found a price that existed only in JSON-LD. ChatGPT search also uses outside search providers, and whether they use your schema isn’t published. Ahrefs’ May 2026 study found no measurable ChatGPT citation lift after pages added schema.

Should I remove FAQ schema from my site?

There’s no need. Google says structured data that isn’t being used doesn’t cause problems, though it has no visible effect either. FAQ rich results no longer show in Google Search from 7 May 2026, so don’t pay for more. Keep visible FAQs if customers find them useful.

Is llms.txt a better bet than schema?

Not for Google. Google says Google Search doesn’t use llms.txt files, and having one will neither help nor harm your visibility there. Other services may use them. For Google, your Business Profile, clear facts on your pages and ordinary rankings matter more than either. Our llms.txt guide covers the rest.