Does Schema Markup Help AI Citations? What the Data Says

TL;DR

Adding schema markup doesn't reliably get you cited by AI. The largest controlled test so far (Ahrefs, 1,885 pages, August 2025 to March 2026) found no meaningful lift in ChatGPT or Google AI Mode and a small drop in AI Overviews. In our own Australian data, cited pages were more likely to have schema than uncited ones (92% against 75%), but that's a correlation. Keep schema for rich results and entity clarity. It isn't a citation lever.

Most advice on schema markup for AI search still ranks FAQPage, Article and Person markup as ways to get cited. The best evidence so far says adding schema doesn't lift AI citations.

This post sets the controlled test against our own Australian data, then checks the usual advice on structured data for AI search, type by type.

What the controlled test found

In May 2026, Ahrefs published the most careful test of this so far. Instead of comparing pages with schema against pages without it, they tracked pages that added JSON-LD and measured what happened to their AI citations.

The method, in plain terms. They took 1,885 pages that added schema between August 2025 and March 2026. They matched them against about 4,000 control pages that didn't. Then they compared citations in the 30 days before and after the change, and measured the difference between the two groups (a difference-in-differences test). That strips out anything that moved both groups, such as an AI engine citing fewer pages overall.

Adding schema didn't lift citations on any of the three engines. On AI Overviews, the treated pages did slightly worse: a 4.6% drop that was statistically significant, though both groups were already falling.

So why did the correlation fool so many people? Schema lives on better-maintained sites. A business that bothers with JSON-LD usually also has a developer, a content process, fresher pages and more authority. Those are the things that earn citations. Schema comes along for the ride.

What Google says

Google's guidance on AI features and your website says there are "no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary." It also says structured data should match the visible content on the page. Google doesn't state that adding schema improves your chances of being cited.

What we found on cited pages

We checked the pages Google's AI Overviews cited for 21 Australian service questions in September 2026, then compared their structured data with pages that ranked on page one but weren't cited.

Page group	Pages checked	Any JSON-LD	FAQPage schema	Author marked up as a Person
All AI Overview-cited pages	116	92%	26%	49%
Cited service-provider pages	82	96%	30%	51%
Page-one pages not cited by AI Overviews	53	75%	4%	Not measured

How we measured. URLs came from Ahrefs SERP data for Google Australia, last crawled September 2026, across mortgage brokers, buyers agents, legal and conveyancing, accountants, dental, builders and SEO. We crawled each page and detected JSON-LD from its HTML, then recorded the schema types present. Forums, video, PDFs and pages that blocked our crawler were excluded. This is a small sample from one engine and one country, so the figures show a pattern rather than a rule.

Schema was common on cited pages, but 75% of pages that ranked on page one without being cited had JSON-LD too. That means schema presence alone doesn't explain the difference. It's a correlation, not a cause.

The FAQPage gap is more interesting: 26% of cited pages had it, compared with just 4% of ranked-but-uncited pages. We read this as a possible signal of question-and-answer content rather than evidence that FAQPage markup gets pages cited. The content may be doing more of the work than the markup itself. That's why the controlled Ahrefs test matters.

The headings point the same way. 47% of cited pages used question-and-answer headings without any FAQPage markup, well above the 26% that had the markup itself. The Q&A format was common on cited pages. The markup was not.

The usual advice against the evidence

Here's the common schema advice for AI search, type by type, alongside what our data and the available evidence show.

Schema type	The usual advice	What the evidence shows	What we recommend now
FAQPage	Add it to get quoted in AI answers	Our data shows FAQPage on 26% of cited pages, compared with 4% of ranked-but-uncited pages. This shows an association, but it doesn't tell us whether the markup itself contributed to the citations or simply reflects the type of content those pages contain.	Write useful Q&A content. Add the markup if it's useful and easy to maintain, but don't expect it to create a citation lift.
Article	Add it so AI trusts your content	We found no evidence that Article markup itself increases citations.	Keep it on relevant articles for structured author and date information.
Person	Add it to prove your author's expertise	49% of cited pages marked the author as a Person.	Put a visible byline and bio on the page first. Make sure the markup matches.
HowTo	Add it for step-by-step answers	HowTo markup wasn't present in our cited-page sample, and Google stopped showing HowTo rich results in 2023.	Focus on clear numbered steps on the page rather than adding it for AI citations.
LocalBusiness	Add it to appear in local AI answers	It can help Google understand business details, but our data doesn't show a citation effect.	Keep it for relevant local businesses and keep the details accurate.

In our audits, schema sits in technical hygiene alongside crawlability and page speed, not in the list of proven AI citation levers. We still check it, fix errors and flag markup that contradicts the page. We don't treat schema as a way to get into AI answers.

What schema is still for

Schema still has useful jobs. They just aren't the AI citation shortcut it's sometimes presented as.

Rich results in classic search. Product details, review information, event dates and breadcrumbs can make eligible pages more useful in Google's standard search results when the markup meets its requirements.

Entity clarity. Organization and Person markup can help Google understand the business and people associated with a page.

Product and local data. For online stores and local businesses, structured data can provide Google with information such as prices, availability, opening hours and locations.

Consistency. Markup should match the visible page content. If your schema says one price and the page says another, fix it. Google specifically asks that structured data match the content users can see.

Set up the relevant markup, keep it accurate and maintain it as the page changes. Then focus your AI SEO effort on the content itself.

What moves citations instead

In our data, the differences that stood out sat on the page, not in the markup. Compared with pages that ranked on page one but weren't cited, cited pages were more likely to name an author (81% against 58%), include a table (32% against 13%) and use question-style headings (70% against 53%). First-party information that isn't available elsewhere gives an AI system something specific to draw on. A clear answer and a specific figure early under a question heading make that information easy to lift.

These aren't individual switches that guarantee a citation. They are patterns that appear across the pages and sources being surfaced, and they give a more useful starting point than treating schema as the deciding factor.

Keep the schema. Stop selling it as a citation fix.

If you're being quoted for a schema project to improve AI citations, ask what evidence supports it. A controlled test found that adding schema didn't increase citations, while correlation alone can't show that schema caused the difference.

Use schema for what it does well, then put your AI SEO budget into content that gives AI clear, useful information to cite.

For more on how AI engines select sources, see how AI search works and why your competitors appear in AI search and you don't.

If you want to see what's holding your pages back in AI answers, start with a free AI visibility audit. Senior-led, two business days, no obligation.

Frequently asked questions

Does ChatGPT read schema markup?

There's no evidence that adding schema changes whether ChatGPT cites you. In the Ahrefs controlled test, pages that added schema saw a 2.2% change in ChatGPT citations, which wasn't distinguishable from zero. OpenAI hasn't publicly explained whether or how structured data affects source selection.

Does schema help you appear in Google AI Overviews?

The current evidence doesn't show that it does. In the Ahrefs study, pages that added schema saw a 4.6% drop in AI Overview citations, while the control group was also falling. Google says there are no additional requirements to appear in AI Overviews.

Which schema types should I still add?

Use Organization markup for your business, and LocalBusiness where relevant. Article markup suits blog posts, and Product markup suits product pages. Add Person markup when there's a real byline and author bio, and make sure the markup matches the visible page content.

Can schema markup hurt AI citations?

There's no clear evidence that schema directly reduces citations. The Ahrefs test did find a small, statistically significant drop in AI Overview citations, but both groups were already falling and the researchers couldn't pin the drop on schema. The clearer risk is inaccurate or contradictory markup. Google asks that structured data match the visible page content.

Is FAQPage schema still worth adding?

Only if it's useful and easy to maintain. Google has limited FAQ rich results to well-known government and health sites since 2023, so most businesses won't see those results. In our data, cited pages were more likely to have FAQPage schema (26% compared with 4%), but we read that as a sign of Q&A content, not proof that the markup caused citations.

Why do so many AI-cited pages have schema if it doesn't cause citations?

Schema is often found alongside other characteristics of well-maintained pages. Sites using JSON-LD may also have fresher content, clearer structure and other signals that help AI systems assess a page. Our data shows the association, but not which factors caused the citations.

What should I do instead of adding more schema?

Focus on useful content that clearly answers the questions your customers ask. It can help to put a direct answer and relevant figures near the top of the page, add first-party information a generic article can't provide, and show a named author and date. You can also look at the forums, directories and comparison sites that AI cites in your category.