Answer engine optimization: be position zero

What an answer engine is, and which ones matter

An answer engine is any search surface that resolves the question on the page instead of handing you a list of links. Google featured snippets and People Also Ask boxes are the oldest and still the largest. Bing has its own answer panel. Voice assistants read one answer aloud. AI summaries sit on top of the same idea and pull from the same kind of source material.

They matter because they sit above the results. A featured snippet occupies the space the first organic result used to own, and People Also Ask boxes often appear halfway down the page, pushing everything below them further out of sight. If a competitor owns the answer box for a question your buyers ask, they get read first even when you outrank them in the list.

The practical point is that these features are populated from pages that already rank. Google draws snippets from the top results for that query, usually the top ten. So AEO is not an alternative to ranking. It is what you do once you rank, to take more of the space.

• Featured snippets: paragraph, list and table formats, drawn from ranking pages
• People Also Ask: expandable related questions, each with its own extracted answer
• Voice results: one answer read aloud, often the snippet source
• AI summaries: same source material, different presentation

How featured snippets and People Also Ask get populated

Google picks a snippet by finding a passage on a ranking page that answers the query cleanly and can be lifted without losing its meaning. The passage has to make sense out of context. If your answer starts with the word therefore, or depends on the sentence above it, it cannot be extracted and something else gets chosen.

The format is chosen by the question. How-to and best-of queries usually produce lists. Comparison and specification queries produce tables. Definition and why queries produce paragraphs. You can see which format Google prefers by searching the query yourself and looking at what is already there. Matching the format that is already winning is the single highest-value move in AEO.

People Also Ask works the same way but at scale. Each question in the box is answered from a page that ranks for that sub-question, and the set of questions expands as you click. It is the cheapest keyword research in existence, because it shows you exactly what related questions Google believes searchers ask next, in their own words.

• Snippets come from pages already ranking, usually inside the top ten
• The winning format is visible in the current result, so match it
• Answers must stand alone with no preceding context needed
• People Also Ask questions are a live map of adjacent intent worth covering

The formats that get extracted, and how to write them

The pattern that works is the same across every answer surface: put the direct answer immediately after the question, then explain. We aim for roughly forty to sixty words for a paragraph answer, written as a complete statement that repeats enough of the question to make sense alone. Then the depth follows underneath, for the reader who wants it.

For list answers, the structure carries the meaning. Use a heading that matches the question, then a real ordered or unordered list with short, parallel items. Each item should be a self-contained step or point, not a sentence fragment that only works in sequence. For comparison questions, an actual HTML table with a clear header row is far easier to extract than the same information written as prose.

Tone matters more than most people expect. A peer-reviewed benchmark and a large correlational study both point the same direction: clear, sourced, non-promotional writing gets picked up more readily than marketing copy, and keyword stuffing performs worse than plain writing. Write it as if you were answering a colleague who has no reason to buy from you.

• A short direct answer of about forty to sixty words, right under the question heading
• Question-shaped headings that mirror how people actually phrase it
• Genuine lists and tables in HTML, not screenshots or styled divs
• Plain, factual tone with sources named where a claim needs one
• One question per section, so the boundary of the answer is obvious

How to structure a page so it can be extracted

Extraction is a parsing problem. The engine has to find where an answer starts and stops, so anything that blurs that boundary hurts. A clean heading hierarchy is the main tool. One h1 for the page topic, h2 for each question, and content that belongs to that question sitting directly beneath it and nowhere else.

Length works against you if it is unstructured. A four thousand word page with no clear sections gives an engine no obvious passage to lift. The same content broken into fifteen question-led sections gives it fifteen candidates. This is why FAQ sections outperform their reputation: they force the structure that extraction needs.

The content also has to be in the raw HTML. The best available evidence on AI crawler behaviour, from a large study of crawler traffic in late 2024, found that several major AI crawlers fetch JavaScript files but never execute them, though Gemini and AppleBot do render. That study has not been replicated, so we treat it as the best current evidence rather than settled fact. Either way, server-rendered answers cost nothing extra and remove the risk.

• One question per heading, with the answer immediately beneath it
• Descriptive headings rather than clever ones, so the topic is unambiguous
• Answers present in the initial HTML, not injected after load
• Accordions that keep their content in the markup, not fetched on click
• Dates, authorship and sources visible, so the passage can be trusted on sight

Where schema markup genuinely helps, and where it does not

Schema markup describes your content in a machine-readable format. It is worth doing, and we do it on every page, but the reason has to be stated accurately. It helps engines resolve what a page is about and which entity it belongs to, and it is required for several rich result types in normal search.

What it does not do is cause AI citation. A difference-in-differences study of 1,885 pages that added JSON-LD, against around four thousand control pages, found no meaningful lift, with a small decline for AI Overviews. Google documentation states there is no special structured data you need to add for its AI features. Anyone selling schema as a route to AI citation is selling something the evidence does not support.

So we use it for what it does: rich results, entity clarity, and removing ambiguity about your organisation, your locations, your products and your authors. FAQ markup still earns its place for the same reason, even though the rich display of FAQ results in Google has narrowed since it launched.

Voice search, honestly

Voice search gets oversold. The realistic picture is that voice queries are longer, more conversational, more likely to be phrased as a full question, and heavily weighted towards local and quick-fact intent. Where do I find, what time does it open, how do I fix, how far is it.

A voice assistant reads a single answer, so there is no second place. That sounds brutal but it changes less than you would think, because the answer usually comes from the same passage that wins the featured snippet. Optimising for extraction covers most of it. What voice adds is a bias towards natural phrasing and towards complete local business information.

For most businesses that means two things. Write headings the way a person would say the question out loud, not the way a keyword tool abbreviates it. And if you serve customers in person, make sure your opening hours, address, phone number and service list are complete and consistent everywhere, because that is what assistants read from.

How AEO depends on SEO, and where the two diverge

AEO cannot rescue a page that does not rank, because the answer features draw from ranking pages. Every technical and authority problem that keeps you off page one keeps you out of the answer box too. In that sense AEO is a second layer applied to pages that have already earned their place.

Where they diverge is in what success looks like. Classic SEO wants the click. AEO sometimes gives the answer away, and the searcher never visits. That is a real trade-off and worth being honest about. For simple factual queries you may win visibility and lose the session. For considered purchases, being the named source of the answer builds the recognition that decides who gets shortlisted later.

The way we resolve it is to choose targets deliberately. We chase answer features on questions where the answer naturally leads to a next step, and where a short summary cannot complete the job. Pricing, process, comparison, eligibility and how-to-choose questions all fit. Pure definition queries we cover for completeness, not for traffic.

• Track snippet and People Also Ask ownership separately from position
• Expect some queries to lose clicks while gaining visibility, and plan for it
• Prioritise questions where the answer creates a reason to keep reading
• Refresh answers when a competitor takes the box, because ownership moves