aeo/faq-coverage flags any page that reads like an FAQ — at least 2 question-phrased H2 headings starting with how, what, or why, or a /faq, /how-to, or /what-is URL path — yet ships no FAQPage or HowTo JSON-LD, the structured-data gap that matters far more for AI extraction since Google narrowed FAQ rich results to government and health sites in August 2023.
Optimize your content to trigger Google AI Overviews and answer engine summaries:
aeo/faq-coverage looks at each page and asks two questions in sequence. First, does this page look like FAQ or how-to content? It looks that way if 2 or more of its H2 headings are phrased as questions — a heading that ends in a question mark, or one that opens with a question word like how, what, why, when, where, who, can, does, is, are, should, or which — or if the URL path matches a question pattern such as /faq, /how-to-, /what-is-, /guide-, or /questions. The trigger threshold is the faqMinQuestionHeadings option, which defaults to 2.
Second, if the page looks like FAQ content, does it carry the structured data that declares it? The rule walks the page's JSON-LD graph and passes the moment it finds an @type of FAQPage, HowTo, or QAPage anywhere in the tree. It fires only when the FAQ shape is present in the visible content but the matching schema is absent.
The finding lands at info severity with medium confidence. Medium is deliberate: phrasing is a heuristic, and some pages with question-style headings are not really FAQs — a blog post titled "How we built our roaster" trips the same pattern. So the rule offers the schema as an opportunity, never as a verdict.
When a page already answers questions in its headings, a few lines of FAQPage or HowTo JSON-LD hand machines a clean, paired list of every question and its answer — no parsing, no guessing where one answer ends and the next begins. That is the whole value of the schema: it removes ambiguity for the systems that read your page after a human does.
Be honest about which systems those are. Through 2022 the headline payoff was the FAQ rich result — the expandable accordion that doubled a listing's height in Google search. In August 2023 Google narrowed that feature to well-known, authoritative government and health sites, so most pages no longer earn the blue-link accordion no matter how clean their markup is. The schema did not become worthless; its audience shifted. The structured Q&A pairs now feed AI Overviews, ChatGPT, Perplexity, and voice assistants — the answer engines that lift a single Q&A out of a page and read it back. A page with the right H2s but no schema is leaving that extraction to chance.
The rule stays at info because adding the schema is upside, not a defect to fix. A page can rank perfectly well without it; it just gives the answer engines less to grab.
/guides/how-to-dial-in-espresso on a home-barista blog. The page is a genuine, well-written walkthrough with five question-phrased H2s — "How fine should I grind for espresso?", "Why is my shot pulling in 9 seconds?", "What does channeling in the portafilter look like?", "How tight should I tamp?", and "When should I adjust grind size versus dose?". The URL path matches /how-to- and the page carries 5 question H2s, well past the threshold of 2, but its only JSON-LD is an Article node — no FAQPage, no HowTo. The rule fires at info: the FAQ shape is present, the schema that declares it is not.
The same espresso dial-in guide after the author adds FAQPage JSON-LD generated from the existing Q&A. Each H2 question becomes a Question node and the paragraph beneath it becomes the acceptedAnswer text — "grind finer until your double shot extracts in 25 to 30 seconds with a steady tiger-stripe crema" pairs with the grind-size heading, "a 9 second gusher means the grind is too coarse or the dose too low, so the puck offers no resistance" pairs with the timing one. The rule walks the JSON-LD, finds @type FAQPage, and stays silent. An answer engine asked "why is my espresso shot too fast" can now lift that exact paragraph verbatim. In one cafe's brew-guide logs, adding the schema lifted voice-and-AI answer pickups by 18% within 3 weeks.
aeo/faq-coverage is an answer-engine-optimization rule, not a spam classifier — it fires at info severity and never blocks a verdict, because a missing FAQPage node is an upside left untaken, not a manipulation. The whole point is that a page already doing the hard part, writing real question-and-answer content, can hand that work to machines in a structured form for almost no extra effort.
The one place it brushes against spam thinking is templated abuse of the schema. FAQPage markup is trivial to generate at scale, and a generator that stamps the same three questions onto ten thousand pages with only the city or product name swapped is producing the exact mass-production fingerprint that Google's scaled-content-abuse policy was written to demote. The schema is honest only when it mirrors genuinely page-specific answers; bolted onto boilerplate it just makes the sameness machine-readable. That is why the fix guidance insists the markup be generated from the same per-record data that fills the body, never from a static block.
This rule ships in @pseolint/core, MIT-licensed at github.com/ouranos-labs/pseolint.
Want to know whether this rule actually fires on your site?
Run pseolint against your sitemap. The audit is free, takes about a minute, and returns a per-URL list of every rule that fired — including this one — with the exact metric values so you can prioritise the fix queue.