Key takeaways
- Match the answer format to the question type: definition, number, list, steps, comparison.
- Lead with a complete 40–60 word answer directly under a question-shaped heading.
- Replace generalities with specifics; answer engines can't lift vague claims.
- Schema must mirror visible content exactly — it confirms the answer, it doesn't replace it.
The one principle
Every answer engine does the same thing: identifies the question, finds the passage that answers it most directly and completely, and shows that passage. Everything in this playbook follows from one principle — make the answer impossible to miss. A human skimming the page should find it in two seconds. So should a machine.
That means a heading that restates the question, a first sentence that answers it, a short complete answer before any elaboration, and a format that matches what the question is asking for. It also means removing the things that get in the way: long preambles, marketing adjectives, answers split across sections, and schema that describes content the page doesn't contain.
Match the format to the question
| Question type | Example | Best format |
|---|---|---|
| Definition | What is generative engine optimisation? | One or two sentences, 40–60 words, then detail |
| Number / fact | How much does technical SEO cost? | The figure or range in the first sentence, then what affects it |
| List | What does an SEO audit include? | Short intro sentence, then a bulleted list of 5–9 items |
| Process | How do I set up Google Business Profile? | Numbered steps, each starting with a verb, with HowTo schema |
| Comparison | AEO vs GEO — what's the difference? | One-sentence distinction, then a two-column table |
| Recommendation | Which CRM is best for a small law firm? | Named options with one line of “best for” each |
Pages that answer a list question in prose, or a comparison in paragraphs, are passed over for pages that use the matching structure — even when the prose is better.
The answer-first page template
This is the structure we use for service pages, guides and FAQ hubs. It works for a single question or a page that answers several.
H1: the main question, or the service named as a buyer would search it
Intro: one paragraph — who this is for and what they'll get
H2: [Question 1, phrased as the buyer would ask it]
Direct answer: 40–60 words, complete on its own
Supporting detail: list / table / steps as the format requires
Specifics: numbers, names, dates, examples
H2: [Question 2]
...
H2: Frequently asked questions
H3 per question, 1–3 sentence answers
FAQ schema matching these exactlyKeep one question per section. If a section answers two things, split it. If two pages answer the same question, merge them or point one at the other — engines won't choose between your own pages for you.
Specificity: the test
Read each answer and ask: could a competitor publish this sentence unchanged? If yes, it's not specific enough to be lifted. Replace adjectives with facts.
- “Fast turnaround” → “Audits are delivered in ten working days.”
- “Comprehensive audit” → “Forty checks across crawling, indexing, rendering, speed, structure and schema.”
- “Proven results” → “218% organic growth in four months for a B2B SaaS client.”
- “Affordable pricing” → “Programmes start at [figure] per month; audits from [figure].”
Specific sentences are also the ones generative engines cite, which is why AEO and GEO are run together.
Schema that confirms, not invents
Structured data tells engines what a page contains. It only helps if it matches what's visible; mismatched schema is ignored or penalised. The useful types for answer content:
- FAQPage — for a set of questions with short answers, all visible on the page.
- HowTo — for numbered processes with distinct steps.
- Article — for guides and whitepapers, with author, date published and date modified.
- Service — for service pages, with provider, area served and description.
- Organization — site-wide, with sameAs links to official profiles.
- BreadcrumbList — to communicate site structure.
Validate with Google's Rich Results Test and the Schema Markup Validator, and keep the schema in sync when content changes.
The pre-publish checklist
- Does the H2 read like the question a buyer would actually ask?
- Is there a complete answer in the first 60 words under it?
- Does the format match the question type?
- Could a competitor publish the answer unchanged? If so, add specifics.
- Is there exactly one page answering this question on the site?
- Does the schema match the visible content exactly?
- Is the page dated, and is the date accurate?
- Does the page load and render without JavaScript?