Schema Markup for AI Search: What It Actually Does
Schema markup for AI search doesn't drive ChatGPT or AI Overview citations the way most guides claim, and a 1,885-page study shows why.
Schema markup does not get a page cited by ChatGPT or AI Overviews; an Ahrefs study of 1,885 pages found adding JSON-LD produced no statistically significant citation lift, because most AI systems read visible page text and ignore the hidden markup.
Type "schema markup for AI search" into Google and you will land on a dozen guides telling you the right JSON-LD will get you quoted by ChatGPT. The claim is everywhere. The evidence for it is not.
What the research actually found
Ahrefs ran the test most of those guides skip: it tracked 1,885 pages that added JSON-LD schema between August 2025 and March 2026, then measured whether citations moved on Google AI Overviews, Google AI Mode, and ChatGPT. The result was no meaningful uplift on any of the three. The ChatGPT effect landed around +2.2 percent, small enough that Ahrefs calls it statistically indistinguishable from zero (Ahrefs, "We Tracked 1,885 Pages Adding Schema. AI Citations Barely Moved.").
The same research explains why. A separate real-time test of how ChatGPT, Claude, Perplexity, Gemini, and Google AI Mode fetch a page found all five extract the visible HTML text and skip the hidden layer entirely. JSON-LD, hidden microdata, and hidden RDFa were ignored across the board. If a model never reads your markup, it can't be swayed by it.
So why do AI-cited pages so often have clean schema sitting in the source? Because the kind of site that bothers to maintain structured data is usually also the kind of site that is well-built and authoritative in the first place, and that underlying quality is what gets it cited. The schema is a symptom of a well-run site, not the cause of the citation.
What schema is still worth doing
None of this means rip your schema out. Google still uses structured data to generate rich results for types like Product, Review and AggregateRating, Article, Recipe, and LocalBusiness (Google Search Central, structured data documentation), and those rich results do affect click-through in regular search. Google itself is also explicit that structured data is not required for AI Overviews or AI Mode: its own documentation states "there's also no special schema.org structured data that you need to add" to appear in those features (Google Search Central, AI features documentation).
Treat schema as plumbing for classic search, not as a GEO lever. If you also want AI engines to quote you, that comes from the visible writing: a self-contained sentence that answers the question with no missing context, under a clear author, on a page that already has some topical standing. For a fuller checklist of what actually moves AI citation, see FAQ Schema in 2026: Still Worth It? and E-E-A-T for AI Engines. If you want a full read on where your own pages stand on this, that is exactly what a GEO audit checks.
The honest version of "schema markup for AI search" is short: add it for Google's rich results, keep it accurate, and stop expecting it to be the reason a model decides to quote you.
Frequently asked
Does adding schema markup help a page get cited by ChatGPT?
Not on its own. Ahrefs tracked 1,885 pages that added JSON-LD schema and found the effect on ChatGPT citations was about +2.2 percent, which the study classifies as statistically indistinguishable from zero.
Why doesn't structured data influence AI citations more?
A separate real-time test cited in the same research checked how ChatGPT, Claude, Perplexity, Gemini, and Google AI Mode actually fetch a page, and found every one of them extracted only the visible HTML content. JSON-LD, hidden microdata, and hidden RDFa were ignored.
If schema doesn't drive citations, why do cited pages so often have it?
Correlation, not causation. Sites that bother to maintain clean structured data tend to also be well-maintained, authoritative sites, and it's that underlying quality and authority that AI engines favor, not the markup itself.
Should I stop adding schema markup?
No. Schema still earns Google rich results for types like Product, Review, Article, and LocalBusiness, and it gives search engines an unambiguous read on what a page is about. It is just not the AI-search lever most guides claim it is.
What actually does influence whether an AI engine cites a page?
The visible content itself: a self-contained answer a model can quote without extra context, clear authorship, and genuine topical authority. Markup can't substitute for content the model can't actually lift a clean answer from.
More in GEO & AI Search
E-E-A-T When the Reader Is a Model
E-E-A-T explained: what Google's experience, expertise, authority, and trust framework actually measures, and whether any of it applies when an AI reads your page.
AI Search Optimization: The Signals That Matter
AI search optimization signals explained: what ChatGPT, Perplexity, and AI Overviews actually look for before citing a page, and how to check for them.
Answer Engine Optimization: Passage-Level Writing
Answer engine optimization means writing self-contained passages, not full pages, so a snippet, voice assistant, or AI engine can quote them directly.
See how AI describes your brand
A free GEO audit takes about 90 seconds and tells you what ChatGPT, Claude, Perplexity and Gemini actually say about you.
Run a free GEO audit →Written by YaBaDo's AI agents and reviewed by a human before publishing. If you find something wrong, tell us and we will correct it.