Do FAQ pages get cited by AI?

A public FAQ label appeared in the page title or address for 26,426 of 7,434,246 cited-page links, or 0.36%. That is about one in 281 links. The links covered 14,251 pages on 8,731 websites.

Selected FAQ pages can earn citations, as Semrush's new-page experiment showed. That does not make FAQ formatting a large source layer across the observed set. Measure FAQ citations against every cited page before treating the format as a general lever.

At least one FAQ-labeled page appeared in 24,070 of 701,353 cited answers, or 3.4%. At least one appeared somewhere in the observed history for 7,602 of 18,107 prompts, or 42.0%.

Prompt reach records whether the format appeared at least once across repeated answers. It does not predict whether the next answer will include one. Report answer reach and prompt reach separately.

ChatGPT Search used 6,812 FAQ-labeled links among 1,663,858 cited-page links, or 0.41%. Google AI Mode used 19,614 of 5,770,388, or 0.34%. The rate ratio was 1.2 to 1, and the gap was 0.07 percentage points.

Keep the engine in the audit, but do not treat this small observed gap as a universal engine preference. The measurement does not control for the engines' different answer and prompt mixes.

The FAQ label appeared in the public address for 22,576 of 26,426 FAQ-labeled links, or 85.4%. It appeared in both title and address for 13,734 links, in the address only for 8,842, and in the title only for 3,850.

A title-only audit misses many publicly labeled FAQ pages. Preserve the cleaned public address with the page title when checking cited pages.

FAQ-labeled pages supplied 2,165 of 93,136 documentation links, or 2.3%. Government websites recorded 1.7%, education 0.73%, ecommerce 0.69%, health 0.62%, and SaaS websites 0.37%. The documentation rate was 6.2 times the SaaS rate.

Compare FAQ pages within a publisher category before setting a target. A general marketing website and a documentation site serve different questions and should not share one benchmark.

Wrike supplied 1,379 FAQ-labeled links, Smytten 617, Mayo Clinic 440, Google 291, Bright Data 234, and Yext 232. The top ten websites together supplied 3,738 of 26,426 FAQ-labeled links, or 14.1%. Wrike alone supplied 5.2%.

There is no single FAQ publisher to copy or target. Build a market-specific list from the exact pages cited for relevant prompts.

Sustainability recorded 343 FAQ-labeled links among 45,957 cited-page links, or 0.75%. Government and military recorded 0.71%, health care 0.57%, transportation 0.54%, education 0.50%, financial services 0.47%, software 0.19%, and clothing and apparel 0.18%. The first rate was 4.3 times the last.

Use an industry baseline before comparing FAQ-page shares. The observed differences describe the prompt and source mix and do not show that industry caused a citation.

Complete FAQ or FAQs tokens supplied 22,062 of 26,426 labeled links, or 83.5%. The phrase frequently asked questions supplied 4,364, or 16.5%. The final links included 1,812 pages without a title, classified from the public address.

The rule misses FAQ content without a public label and does not read page content or structured data. It measures labeled pages, not format quality, traffic, causal lift, or whether the page changed an AI recommendation.

FAQ-labeled pages supplied 26,426 of 7,434,246 cited-page links. Documentation websites had a 2.3% rate, SaaS websites had a 0.37% rate, and the top ten websites supplied 14.1% of labeled links.

Keep the cited URL, page title, engine, prompt, website category, and industry together. Audit the buyer questions where an FAQ answer is useful. Compare against the correct publisher and industry baseline. Do not infer citation lift from adding an FAQ label, format, or schema. Rerun the fixed method next quarter before calling any difference movement.