Skip to main content
AI answers cite content that’s already formatted like an answer — discrete question-and-answer blocks, accordion patterns, headings that pose the question your page resolves. Pages that hide answers inside long unstructured prose get passed over for sites that format for the cite.

​
Methodology

Cited samples up to 5 pages from your site and looks for three structural patterns AI models recognize as answer blocks. Each pattern is worth points; you don’t need all three to score well. FAQ structures. The strongest signal. We detect FAQ content in three ways: JSON-LD FAQPage schema (the explicit machine-readable declaration), HTML elements with class or id matching faq or question whose text content contains a ?, and <details><summary> accordion pairs. Any page with at least one FAQ structure earns 4 points — this is the single largest contributor to the score because FAQ schema produces direct AI answer citations more reliably than any other pattern. Accordion elements. The <details> and <summary> HTML pair is a native browser accordion that AI parsers treat as an implicit Q&A structure. We count <details> elements separately from FAQ detection because some sites use accordions for non-FAQ content (product specs, terms-and-conditions). Pages with accordions earn 2 points if the FAQ check hasn’t already maxed that tier. Question-style headings. H2 and H3 headings that start with a question word — what, how, why, when, where, who, which, can I, do I, does, is there, is it, should, will, are — signal to AI parsers that the content under each heading is an answer to that question. Without FAQ or accordion structures, three or more question-style headings on a page earns 3 points; five or more earns 4. When FAQ or accordion structures exist, question headings earn 1 bonus point. Multi-page consistency bonus. Sites with answer-block patterns on more than one sampled page earn an additional 1 point. AI models cite sites that consistently format for citation; one-off FAQ pages signal less authority than sitewide structure. The signal scores out of 7. Sites with strong FAQ schema across multiple pages plus question-style headings typically score 6-7/7. Sites with one FAQ page and no other structure score 4-5/7. Sites with only prose content and no question patterns score 0/7 and are flagged.

​
Verification

You can verify our finding yourself in a browser. Step 1: Open the pages we sampled. Cited reports the URLs we tested. Open each in a new tab. Step 2: Check for FAQ schema. Right-click the page, View Page Source, and search (Cmd+F / Ctrl+F) for application/ld+json. Inside any <script type="application/ld+json"> block, look for "@type": "FAQPage". If present, the page declares FAQ content explicitly — this is the strongest signal. Step 3: Check for accordion elements. In the page source, search for <details> and <summary>. Every <details> element produces a native browser accordion. If you find several, your page has implicit Q&A structure. Step 4: Check for question headings. Open DevTools (Cmd+Option+I / Ctrl+Shift+I), Console tab, and run: 
[...document.querySelectorAll('h2,h3')].filter(h => /^(what|how|why|when|where|who|which|can\s+i|do\s+i|does|is\s+there|is\s+it|should|will|are)\b/i.test(h.textContent.trim())).map(h => `${h.tagName}: ${h.textContent.trim()}`)
This returns every H2 or H3 that starts with a question word — the same set Cited counts. If you find three or more, you have a meaningful question-heading pattern; five or more is excellent. Step 5: Validate FAQ schema with Google’s tool. Paste the page URL into Google’s Rich Results Test. If the test detects FAQPage schema, AI models that consume Google’s index will detect it too. If your verification disagrees with Cited’s finding, that’s a bug — let us know.

​
Technical detail

FAQ structures are recognized by AI models through three layered signals: explicit schema declarations (Schema.org FAQPage, defined in 2017 and continuously refined), HTML semantic elements (<details> / <summary> introduced in HTML5, 2014), and pattern-based content extraction (question-form headings paired with answer paragraphs). Detection logic. Cited’s scanner runs three parallel detection passes per sampled page:
  • JSON-LD FAQPage parsing. Every <script type="application/ld+json"> element is parsed. Blocks with top-level @type: "FAQPage" (or arrays containing one) are extracted. Each mainEntity item with a non-empty name field is recorded as a Q&A pair.
  • FAQ-class element scanning. Elements matching the CSS selectors [class*="faq" i], [id*="faq" i], [class*="question" i], or [id*="question" i] (all case-insensitive) are inspected. Any element whose text content includes a ? is recorded as a question; the next sibling element’s text is recorded as the answer.
  • <details> / <summary> extraction. Every <details> element with a <summary> child is recorded. The summary becomes the question; the remaining children become the answer. Separately, the total count of <details> elements is captured for the accordion check.
Question-heading detection. The scanner applies a regex against H2 and H3 headings: /^(what|how|why|when|where|who|which|can\s+i|do\s+i|does|is\s+there|is\s+it|should|will|are)\b/i. Case-insensitive, word-boundary anchored, matches the heading’s text content after trimming whitespace. Score combination. The three patterns aren’t strictly additive — the scoring logic prioritizes FAQ presence (worth 4 of the 7 points), then accordions (worth 2 if FAQ didn’t max), then question headings (worth 2-4 standalone or +1 as a bonus when other patterns exist), plus a 1-point multi-page bonus. The total is clamped to 7. Edge cases the scanner handles:
  • Malformed JSON-LD. Schema blocks that fail to parse (syntax errors, truncated content) are silently skipped. The other detection passes still run.
  • Nested FAQPage in an array. Some CMSes wrap schema in an array. The scanner checks both data["@type"] === "FAQPage" and Array.isArray(data) && data.find(d => d["@type"] === "FAQPage").
  • FAQ-class elements without ?. A <div class="faq-section"> containing only a heading and intro text doesn’t count — the ? filter ensures we only capture actual questions, not section wrappers.
  • <details> without <summary>. Rare but technically valid HTML. These contribute to the count but not to the FAQ list because there’s no question text to extract.
  • Question headings inside FAQ structures. A page with explicit FAQPage schema AND question-style H2s gets credit for both — the FAQ count drives the primary score, and the question headings add the +1 bonus.
What this signal does not measure:
  • Answer quality. The scanner detects that a page has Q&A structure, not whether the answers are good. A FAQPage with three useless one-line answers passes the same as one with detailed answers.
  • Answer length. Some AI citation studies suggest answers between 40 and 80 words cite most reliably. The scanner doesn’t model this; it counts presence.
  • Question-answer pairing accuracy. For the next-sibling FAQ-class detection, the scanner assumes the element following a question is the answer. Real sites sometimes have wrapper divs or styling elements between them; in those cases the extracted “answer” may be a wrapper.
  • Non-English question patterns. The question-word regex is English-only. Pages in other languages with valid question structures score lower on the question-heading check, but FAQ schema and accordions still detect correctly.
For brands without any answer-block structure today, the highest-leverage fix is adding FAQ schema to one cornerstone page — the page you most want AI to cite when users ask about your category. This earns 4 points immediately and signals to AI that the page is citation-ready. See also: Heading Hierarchy, Content Depth.