An Answer-First Page Structure for AI Search
Put a complete answer near the top, under a heading that matches the question, and write it so it still makes sense if quoted on its own. Treat that as an editorial habit rather than an engine rule: no assistant publishes a word count it rewards, and Google's own guidance for its generative features says you can ignore tactics like chunking content, because its systems can find the relevant part of a longer page. The reason to do it anyway is that it helps the reader first, costs nothing, and makes your answer easier for anyone — person or system — to quote accurately. Then structure the rest as one question per heading with its answer immediately below. The common failure is the warm-up intro: two paragraphs of context before the answer.
This is not a claim that structure gets you cited — plenty of well-structured pages are never quoted, and being retrievable is a precondition rather than a cause. What structure does is remove a specific, avoidable reason a page that deserved to be cited was not.
Write answers that survive being quoted alone
Whatever a given system does internally — and most do not publish it — a quoted sentence arrives without the paragraphs around it. A passage that depends on three paragraphs above it for its meaning loses that meaning when lifted out. So write the sentences that carry the answer to stand alone, which mostly means naming the subject rather than relying on a pronoun. This is about quotation accuracy, not about gaming how pages are split up: Google says explicitly that there is no requirement to break content into small pieces for its AI features.
The structure
| Element | Rule | Why |
|---|---|---|
| H1 | The question or the subject, plainly | Sets what the page is about |
| First paragraph | A complete answer, as short as it can be while complete | Readers get the answer without scrolling |
| Each H2 | One question, phrased as someone would ask it | Makes each section easy to find and quote |
| Under each H2 | The answer immediately, then support | Answer-then-evidence, not evidence-then-answer |
| Tables | Labelled columns, one fact per cell | Extractable without prose reconstruction |
| Dates | Stated in text, not only in metadata | Survives extraction |
Write the answer so it can be quoted against you
A passage that is hedged into meaninglessness does not get quoted, and a passage that overclaims gets quoted and then makes you wrong in public. The useful test is whether you would be comfortable seeing the sentence attributed to you, out of context, with no surrounding qualification — because that is the form it will appear in if it appears at all.
What this does not fix
Structure does not make a thin page substantial, and it does not make an assistant prefer you over a source with more authority on the topic. It is a hygiene layer. If the honest state of the page is that it repeats what is already well covered elsewhere, restructuring it is rearranging the problem. Our refresh-versus-new-page guide is the better starting point in that case.
FAQ
How long should the opening answer be? — As long as it takes to be complete and no longer. A short paragraph is a useful editorial target — many writers land somewhere around 60 to 80 words — but that is a habit, not a threshold any engine is known to enforce. We have no evidence that a longer opening gets truncated or penalised, and an earlier version of this page implied otherwise.
Does this hurt readability for humans? — Generally no. Answer-first is standard practice in technical writing and news because readers also benefit from not waiting. What it does cost is the essayistic opening, which is a real loss on some kinds of writing and not on a reference page.
Should I chunk my content for AI? — Not for Google: its guidance for generative Search features says there is no requirement to break content into small pieces, because its systems can identify the relevant part of a page. Other assistants do not publish equivalent guidance. Clearly structured pages written for readers leave nothing on the table.
Do FAQ sections still help? — As structure, yes, because one question per heading with its answer below is the pattern this whole guide describes. As a schema trick for rich results, treat that separately and check current documentation rather than assuming.
Put this to work on your own website.
Fenn finds what your customers ask, drafts the articles and site fixes, and measures what ChatGPT, Claude, Gemini, Perplexity and Grok say about you — with every change waiting for your approval.