Do FAQ pages help you get cited by AI search engines?

Does an FAQ format make AI engines use your page more?

No: in the largest study on this, question-and-answer pages were used slightly less than other pages. The researchers ran 602 test questions through ChatGPT, Google’s AI Overview and Gemini, and Perplexity, then examined 18,151 cited pages they could fetch (Zhang Kai and colleagues (opens in a new tab)).

They separated two outcomes. Being cited means a page appears in the source list. Being used, which they call absorption, means the answer’s wording and facts come from that page.

They scored use by how often and how early a page was referenced and how much of the answer overlapped with it. On that score, pages in a question-and-answer format were used 5.74% less on average than other pages.

The authors are careful: this was a simple average with no controls, and their label for FAQ pages may be noisy. They say it “does not prove that FAQ content is harmful.” It does show the format alone is no shortcut.

What do AI engines actually draw on, if not the format?

They draw on the evidence inside a page. In the same study, pages with certain kinds of content were used far more than pages without them:

The most-used pages were also built in sections. Among the quarter of pages that answers drew on most, the average page had 10.59 headings, against 0.85 for the quarter drawn on least.

The authors’ explanation is plain. An FAQ often produces short, isolated answers that cannot support a longer response. A detailed explainer gives the engine more to work with.

Their advice is that the value “comes from the evidence inside the page, not from the presence of question marks in headings.” The wider pattern is laid out in our guide to what content AI engines favor.

Do FAQ headings or FAQ markup help in Google’s AI Overviews?

The evidence says no reliable lift. In our AI Overview study, we compared 3,096 pages ranking in Google’s top 10 for US searches. Every search showed an AI Overview, the AI summary at the top of Google’s results.

We compared each page only with pages on the same search, so differences in topic did not count. On that basis, a page with an FAQ heading was 2.5 points less likely to be cited, a gap we could not tell apart from zero.

Question-style headings looked like a strong signal in our first version, at 9.6 points. Compared like with like, that fell to 3.5 points and is no longer distinguishable from zero.

FAQ markup, the code that labels questions and answers for search engines, showed a 5.9-point gap. It did not hold up once we allowed for testing 14 page features at once. Ranking position mattered far more than any of these features.

Does adding an FAQ change anything in controlled experiments?

Barely. In a lab test built on 13,747 shopping questions paired with Amazon listings, researchers rewrote listings 15 different ways. They measured how far each version moved in five AI engines’ product rankings (Bagga and colleagues (opens in a new tab)).

Adding an FAQ was one of only four rewrites that matched or beat a plain rewrite. Its average gain was 0.05 places, close to nothing. The other eleven rewrites did worse, and some, like making listings read like advertisements, did real harm.

A second lab test points the same way. Researchers at the software company Sprinklr ran 252,000 trials across six AI models, each time giving the model two near-identical sources (Vishwakarma and colleagues (opens in a new tab)).

Changing layout alone, such as dense paragraphs against organized sections, had no consistent effect on which source was cited first. Relevance, prices and dates did. The split evidence on layout changes is weighed in our guide to restructuring existing content.

When does an FAQ page actually work?

An FAQ works when it is deep enough to answer real questions. The clearest example comes from a study of hotel recommendations in Tokyo, run by an AI company’s researchers (Zhu and Chang (opens in a new tab)).

They scored 14 hotel websites on content depth, including how full their FAQ was. Hotels that Gemini cited directly averaged 8.6 out of 15 on depth; hotels it did not cite averaged 3.4.

One independent hotel, Kadoya, had no special markup but a 33-question FAQ and a detailed sightseeing guide, and Gemini cited it directly. Another, Hotel K5, had a 9.6 out of 10 rating on a booking site, markup and an FAQ, but each feature was brief. Gemini named K5 but drew its facts from booking and editorial sites instead.

This is 14 hotels, one city and one engine, and the authors say it shows association, not cause. It fits the larger studies: the depth of the answers mattered, not the FAQ label. The full hotel comparison is in our guide to deep content without technical SEO.

What should you do about it?

Treat an FAQ as a container for evidence, not as a ranking tactic.

1. Do not convert pages to FAQ format expecting AI engines to use them more. No study we reviewed shows that the format alone does this.
2. Where you keep an FAQ, make each answer substantial. Include the specific figure, the definition, the comparison or the steps a buyer needs.
3. Put your strongest evidence where it is easy to lift: clear sections with headings, numbers stated plainly, and comparisons laid out side by side.
4. Keep markup as tidy housekeeping, not as a growth plan. In our data it did not reliably move citations.
5. Measure whether answers use your page, not only whether they link to it. The two can differ widely.

If you want help turning existing pages into evidence AI answers can use, see our generative engine optimization service.

What does the research not tell us yet?

No study has yet tested FAQ formatting on live pages with a proper before-and-after design.

• The main FAQ result is a simple average. It does not rule out that thin FAQ pages, not the format, explain the gap.
• Our AI Overview study and the hotel study observed pages as they were. Neither changed a page and watched the result.
• The lab tests used product listings and paired sources, not full websites in live AI search.
• Results come from 2025 and 2026 versions of fast-changing products. Engines may weigh pages differently after updates.
• Nobody has measured whether FAQ pages drive more clicks or sales from AI answers.

Frequently asked questions

Should we add FAQ schema to get into AI Overviews?

Not as a growth tactic. In our study of 3,096 ranking pages, FAQ markup showed a 5.9-point gap that did not survive allowing for testing 14 features at once.

Is FAQ content bad for AI search?

No. The largest study found question-and-answer pages used 5.74% less on average, and its authors say this does not prove FAQ content is harmful.

What content do AI answers draw on most?

Pages with numbers, definitions, comparisons and steps. Pages with numbers or statistics were used 61.55% more than pages without them in a study of three AI search engines.

Do question-style headings help a page get cited?

The evidence is weak. In our data the gap fell from 9.6 points to 3.5 once pages were compared on the same search, and it could not be told apart from zero.

Sources

• Zhang Kai, He Xinyue and Yao Jingang (2026), From Citation Selection to Citation Absorption: A Measurement Framework for Generative Engine Optimization Across AI Search Platforms (opens in a new tab), arXiv:2604.25707.
• Bagga, Farias, Korkotashvili, Peng and Wu (2026), E-GEO: A Testbed for Generative Engine Optimization in E-Commerce (opens in a new tab), arXiv:2511.20867.
• Vishwakarma, Kumar and Jamidar (2026), What Gets Cited: Competitive GEO in AI Answer Engines (opens in a new tab), arXiv:2605.25517.
• Zhu and Chang (2026), The End of Rented Discovery: How AI Search Redistributes Power Between Hotels and Intermediaries (opens in a new tab), arXiv:2603.20062.
• Underneath (2026), What pages cited by AI Overviews have in common: 3,096 pages