Does Schema markup (JSON-LD) help a site get cited by ChatGPT or Google AI Overviews?

According to an Ahrefs study tracking 1,885 pages that added Schema markup against 4,000 control pages, adding it produced no significant citation increase on any platform - not Google AI Overviews, not AI Mode, not ChatGPT. If a page is already being crawled and cited, schema doesn't push it higher. It may only help brand-new pages that aren't being crawled at all yet.

So what actually drives citation likelihood in AI engines?

The structure of the content itself: a direct, clear answer at the very top of the page, headings that map a question to its answer, and concrete information (numbers, dates, sources) phrased as a standalone, quotable sentence. A study on ChatGPT citations found that 44.2% of citations come from the first 30% of a document - meaning where the answer sits in the document matters more than how it's technically tagged.

How many Google searches already show AI Overviews?

Per a Semrush analysis of over 600,000 keywords, AI Overviews appeared in roughly 16% of searches as of November 2025, after peaking around 25% in July of that year. The exact figure varies by study depending on measurement method, but the trend is clear: a growing share of searches get a direct answer before the user ever reaches a website.

← All Articles

GEO/AEO · Aug 5, 2026 · May I

What Actually Gets Your Site Cited by AI - and Why Schema Markup Alone Isn't Enough

Short answer: adding Schema markup (JSON-LD) to a page barely moves the needle on whether it gets cited by ChatGPT, Perplexity, or Google AI Overviews. An Ahrefs study tracking 1,885 pages that added schema markup against a matched control group found citations barely changed. What actually decides it is the structure of the content itself - a direct answer at the top of the page, headings that map a question to its answer, and real numbers backed by a source. If you invested in technical tagging and expected results, this is likely why you didn't see any.

The expectation vs. the reality

Over the past year, almost every GEO (Generative Engine Optimization) guide online recommends the same thing: add FAQPage schema, add BreadcrumbList, tag everything in JSON-LD, and AI engines will start citing you. The logic sounds reasonable - schema is the language machines "understand," so more tagging should mean more understanding.

The reality, based on real data, is different. Ahrefs tracked 1,885 pages that added schema markup between August 2025 and March 2026, comparing them against roughly 4,000 control pages at similar citation levels that never added tagging. The result: Google AI Overviews citations actually dropped 4.6% (statistically significant), AI Mode was up 2.4%, and ChatGPT was up 2.2% - the latter two not statistically significant. In other words, adding schema to an existing page did not push it toward more citations.

Why this happens - what AI engines actually "read"

There's a technical explanation too. A companion experiment by searchVIU tested whether five major AI tools - ChatGPT, Claude, Perplexity, Gemini, and Google AI Mode - actually use schema markup when fetching a page in real time. The answer: none of them did. These tools read the visible content, not the structured tagging sitting underneath it.

That doesn't mean schema is entirely useless - it may still help brand-new pages that aren't being crawled at all get indexed. But if the goal is actual citations, it isn't the lever to pull.

What actually decides it: where the answer sits, not how it's tagged

A study by Kevin Indig published on Search Engine Land looked at ChatGPT citations and found that 44.2% of citations come from the first 30% of a document (source: searchengineland.com). In other words, AI engines don't "read everything and then decide" - they disproportionately favor content that appears early. A page that buries its direct answer in the fifth paragraph, after a long intro about "who we are" and "why this matters," simply loses citations to a competitor saying the same thing in the first sentence.

That's why every article we write at May I - including this one - gives the full, direct answer in the opening paragraph, before any context or backstory. It reads less "literary," but it's what actually gets crawled and cited.

Why this matters for your business

Per a Semrush analysis of over 600,000 keywords, AI Overviews appeared in roughly 16% of Google searches as of November 2025, after peaking around 25% in July of that year (source: semrush.com). The exact number varies between studies depending on measurement method, but the direction is clear: a growing share of searches get a direct answer inside the engine itself, before the user ever clicks a link. A site that isn't phrased to be quotable simply doesn't enter that part of the search - no matter how much schema it has.

What to do instead of focusing on tagging

Open every page and article with the direct, full answer - before context, story, or an "about us" preamble.
Use headings (H2/H3) that are themselves the questions someone would type into a search engine.
Back every claim that can carry a real number with one, linked to a verified source - not estimates, not "in our experience."
Schema is still worth adding as sound technical hygiene - just not as a substitute for good content structure.

Want a site built for the AI era, not just Google circa 2015?

We build websites and content with a structure designed to be read and cited by search engines and AI tools alike - not just technical tagging that doesn't actually do the job.

Get started