What this check tests
A crawled page (the homepage plus up to nine more) counts as FAQ content when at least one of these is true of its HTML:
- An element with an
itemtypecontainingFAQPage, which is microdata. - Two or more
<details>elements. - Two or more
<dt>elements. - The visible text contains "frequently asked", in any case.
- The visible text has three or more places where a question mark is followed by whitespace.
The score is 33 points per counted page, capped at 100: one page gives 33, two give 66, three give 99 and four or more 100. Any score above 0 is a pass; no page is a fail. When the check fails, the report offers a JSON-LD FAQPage snippet.
Note the mismatch in that last point. A JSON-LD block of type FAQPage is not one of the signals. It helps only if the same page also shows one of the visible patterns, for example because the questions appear in the text with question marks. The last signal is also broad: ordinary prose with three rhetorical questions, or a list of search suggestions, will count. A page that is not an FAQ can pass, and a page with a real FAQ written as headings and paragraphs, with fewer than three question marks, can fail.
Why it matters to agents
schema.org defines FAQPage as a web page presenting one or more frequently asked questions. That is a content type, not a promise of any display.
Google ended its special treatment of the markup. Its documentation updates record that the FAQ rich result feature is no longer shown in Google Search results, and that the documentation for it was removed. Google's guide to generative AI search adds that no special schema.org markup is needed for AI features and that you do not need to write in a particular way for them. No source fetched says an assistant prefers FAQ sections.
What remains is the ordinary case for an FAQ: people ask the same questions, and a page that answers them clearly saves them a search. This check weighs 0.5, the joint lowest weight in its category, and the sources above do not justify more.
How to fix it
Only add an FAQ if you have real, recurring questions with useful answers:
- Write the question and answer as visible content, on the page where the question comes up.
<details>and<summary>works well for a list that can be expanded, and headings with paragraphs work for a few:
<details>
<summary>Can I change my plan later?</summary>
<p>Yes. Plan changes take effect at the start of the next billing period.</p>
</details>
- Use two or more
<details>so the audit sees it, or put "Frequently asked questions" in a heading. - If you also add
FAQPageJSON-LD, make it match the visible questions exactly. Google's policy says not to mark up content that is not visible to readers. Check the block with schema.org validity; for the types you can use, see schema.org type. - Do not generate questions to fill a quota. One page with a real FAQ is enough to pass, and 33 points are not a goal in themselves.
Related: answer-first content and citation readiness, which carry the same caveat about evidence, and the schema.org for AI background page.