Skip to content
SilktideHelp

Answer-first page structure

An answer-first page states its point - the answer, the offer, the conclusion - in the first paragraph, then supports it, with every section written so it still makes sense when lifted out alone. It is the inverted pyramid journalists have used for over a century, applied to the way pages are actually consumed now: by skimming humans and by machines assembling answers from fragments.

Throughout this page, suppose you run Fernwood, an expense-management product, and the page in question answers "how long do expense reimbursements take?"

Why it works

Three audiences read your page, and none of them reads it top to bottom:

  • AI assistants extract, they don't read. When an assistant answers a question, it retrieves candidate pages, pulls the passages that look like answers, and quotes or paraphrases them. A page whose answer is stated cleanly in one place is easy to quote correctly; a page whose answer is smeared across twelve paragraphs gets skipped for a competitor's page - or worse, gets paraphrased wrongly. Retrieval also works on chunks: long pages are split into passages and each passage is judged alone, so a section that only makes sense in context loses its meaning at exactly the moment it is being evaluated.
  • Search engines rank passages. The heritage here is the featured snippet: Google has long extracted direct answers from within pages, favouring passages that answer a question compactly. Its guidance on creating helpful content asks the same question this technique does: does the page deliver what the visitor came for, without making them work for it?
  • Humans scan. Nielsen Norman Group's eyetracking research shows readers scan in an F-shaped pattern, reading the opening and the left edge - first lines of paragraphs, headings - and skipping most of the rest. Whatever is not at the top or in a heading, most visitors never see. This is also why answer-first pages convert: the visitor who got their answer immediately trusts you with the next question.

The uncomfortable implication: the traditional marketing instinct - build anticipation, establish context, reveal the point - actively fights all three audiences.

Anatomy of an answer-first page

1. A title that matches the question

The page answering "how long do expense reimbursements take?" should be titled "How long do expense reimbursements take?" - not "Getting paid back, faster." Retrieval starts with matching the query; a clever title matches nothing. This applies to the page's <title> and its H1 alike.

2. The answer, in the first paragraph

Open with a self-contained statement of the answer, roughly 40–60 words, that would be accurate and complete if quoted with nothing around it:

Fernwood reimburses approved expenses in the next payment run after approval. Payment runs happen every business day for direct deposit (funds arrive in 1–2 business days) and weekly for cheques. An expense submitted Monday and approved Tuesday is typically paid by Thursday.

Compare the opener this replaces:

We know how frustrating it can be to wait for money you're owed. At Fernwood, we've always believed reimbursement should be painless, which is why we've invested heavily in our payments infrastructure over the years...

The second tells the reader (and the assistant) nothing extractable. Write the first paragraph as the quote you want repeated - because it is what will be repeated.

3. Headings that are sub-questions

After the answer, each H2 takes one follow-up question a real reader has, phrased the way they would ask it: "What can delay a reimbursement?", "How do international reimbursements work?", "When are payment runs?" Headings phrased as questions do double duty - they are scannable signposts for humans and query matches for machines. Keep the hierarchy honest: H2s for sub-questions, H3s only for their parts, no skipped levels.

4. Sections that survive excerpting

Write each section as if it might be the only part of the page anyone sees - because in retrieval, it often is:

  • Re-name, don't just refer. A section that opens "This can also affect it" is meaningless out of context. Open with "International reimbursements can take longer" - the noun restated, the pronoun gone.
  • One answer per section. When a section answers two questions, split it; each becomes independently findable.
  • Front-load each section the same way you front-loaded the page: first sentence is the section's answer, detail follows.

5. Extractable structures for enumerable facts

Steps go in numbered lists, options and comparisons in tables, everything else in short paragraphs. Machines lift list items and table rows verbatim and reliably; the same facts written as flowing prose get paraphrased, and paraphrase is where errors enter. If the fact has units, numbers, or a yes/no shape, put it in a structure that preserves it.

6. Specifics, everywhere

"Fast reimbursements" is unquotable marketing; "paid in the next daily payment run" is an answer. Concrete numbers, named limits, and dates are what make a passage worth extracting - and they are what Silktide's AEO reports look for when judging whether a page is specific rather than promotional. Pair every claim with its evidence, and declare when the page was last updated in metadata - see Machine-readable dates - so an assistant can trust the specifics are current.

What this is not

  • Not question-heading theatre. Pages stuffed with question headings whose sections never answer them ("How long do reimbursements take?" followed by three paragraphs about company values) are worse than unstructured prose - the mismatch between promise and content is exactly what both Google's helpful-content guidance and answer engines discount.
  • Not the recipe-blog anti-pattern in reverse. The 900-word personal story before the recipe is the canonical answer-buried page. But amputating all depth is the opposite failure: answer-first means the answer comes first, not that the evidence, edge cases, and detail get deleted. Depth below the answer is what makes the page worth citing over a one-line competitor.
  • Not a template for padding. Adding an "In conclusion" restatement, a generated FAQ that re-asks the H2s, and three filler paragraphs to hit a word count makes the page worse at everything this technique is for. If the honest answer is four sentences, publish four sentences.

How Silktide helps

Silktide's AEO reports judge exactly this technique, page by page. Four of the report's checks are its core:

A flagged page in the report is a page where this technique has not been applied; this page is how you apply it.

Last updated

Was this page helpful?