What is answer engine optimization?

The short version

Answer engine optimization is the work of being the source an engine uses when it answers a question directly instead of returning links. It is older than the current wave — featured snippets and voice assistants were answer engines too — but the stakes changed when answers stopped citing anyone by default and started naming two or three businesses instead of ten.

The hard part is that a mention often arrives with no link, so the event you care about happens somewhere you have no logs. That is why AEO has to be measured by asking, not by reading analytics.

What counts as an answer engine?

Anything that replies rather than lists:

• assistants asked directly — ChatGPT, Claude, Gemini, Perplexity;
• AI Overviews at the top of a Google results page;
• voice assistants, which have worked this way for a decade;
• featured snippets, the original answer engine;
• in-product search that summarises rather than ranks.

They differ in one way that matters commercially: some cite sources and some do not. Perplexity cites heavily and visibly. A chat assistant answering from memory may name your business without linking to anything at all — good for you, and invisible in your reporting.

Why “no link” is the whole problem

A featured snippet took your click but gave you attribution. An answer engine often gives attribution without a click, or a click without attribution, and sometimes neither.

That breaks the habit of measuring visibility with an analytics tool. Consider what actually happens: a buyer asks an assistant who to use, reads three names, remembers one, and searches for it directly a day later. Your analytics records a direct visit with no referrer and no explanation. The decisive moment — being named — left no trace on your property at all.

The only way to see it is to ask the questions yourself, repeatedly, and record what came back. This is not a preference for a particular tool; it is the only place the data exists.

How do answer engines choose a source?

Nobody outside the companies building them knows the full answer, and anyone claiming otherwise is guessing. What is observable:

They prefer passages that stand alone. An extracted answer travels without its surroundings. A paragraph that only makes sense after three paragraphs of context is a poor candidate, however well written.

They cross-check facts. Where your own site, three directories and a review platform describe your services differently, there is less reason to rely on any of them.

They read more than you think. In one local market we measured, a single answer rested on eleven pages on average — the businesses’ own pages alongside directories, review sites and local roundups. Your pages are the ones setting the facts the rest repeat, which is why they are where the work starts.

They disagree with each other. Four assistants asked the same questions in that market named 63 businesses between them; only four were named by all four, and 46 appeared in a single assistant only.

What does AEO work consist of?

Answer the question first, plainly

The inverted pyramid is not a style preference here. An engine extracting an answer takes the first self-contained sentences that address the question. Burying the answer under context loses it.

Practically: the first two sentences under a heading should answer the heading. Everything else is supporting detail for the human who stayed.

Make each section self-sufficient

Define terms in place rather than linking to a glossary. Repeat the noun rather than writing “it” across a paragraph break. An extracted passage has no antecedents to resolve.

Use the question as the heading

Not Pricing but How much does an emergency appointment cost? — the shape a buyer asks in. This maps directly onto how people prompt assistants, and it makes the passage easier to match to a question.

Add structure a machine can parse

Lists, tables as HTML rather than images, and consistent heading levels. Schema markup where it describes something genuinely on the page — and not where it does not, because marked-up content a human cannot see is a penalty waiting to happen.

Keep facts consistent everywhere

This matters more than any single page, and it is the least interesting work in the list. Services, hours, address, specialities, prices: identical on your site and on every directory that describes you.

Publish something checkable

Specific numbers, stated method, real examples, dates. Engines prefer a source that says something specific to one that says something impressive, and so do the people who read the answer.

How do you measure AEO?

Ask the buying questions. Written the way buyers ask, grouped by intent — general recommendation, service, price, urgency, trust, comparison. Visibility is rarely uniform across those groups, and the group you are absent from is the work.

Ask each one several times, per engine. Answers vary between identical asks; five repeats is our default. One check is an anecdote.

Record three things: whether you were named, who was named instead, and what was cited. The second and third are more actionable than the first.

Never average across engines. A blended number hides the engine you are absent from, which is the one worth knowing about.

The method is published in full, and the sample report shows the output with no form in front of it.

AEO and GEO: is there a difference?

Mostly framing. AEO emphasises the direct-answer format and its older surfaces; GEO emphasises generated text and the newer ones. The work overlaps almost completely, and the same measurement covers both.

If a supplier presents them as two services with two price tags, that is a pricing decision rather than a technical one. The longer version is in generative engine optimization and GEO vs SEO.

What AEO cannot promise

Placement. The lists are not published, not stable and not for sale.

Traffic. An answer engine that answers the question fully reduces clicks even when it names you. Being absent is still worse, but if your only success measure is sessions, this work will look like a failure while succeeding.

Speed. Your own pages move as fast as you approve them; the listings that repeat them move at their moderation queue’s pace, not yours.

Certainty about cause. We can show what changed and what moved. Putting dates next to each other is honest; claiming causation is not.

Common mistakes

Writing for the engine instead of the buyer. Pages stuffed with question headings and no substance get extracted once and then stop.

Marking up what is not on the page. FAQ schema for questions a human cannot see is the fastest way to lose rich results entirely.

Measuring one engine and calling it AI visibility. It describes roughly a quarter of the picture.

Treating a single good answer as proof. Ask again tomorrow.

Which pages should you make answerable first?

Not all of them, and not the ones you are proudest of.

Start with the pages that match a buying question you are absent from. If the measurement shows you never appear in price answers, the page to rewrite is the one about prices — and the rewrite is usually structural rather than stylistic: move the answer to the top, use the buyer’s phrasing as the heading, state the number.

Then the pages a buyer reaches after the answer. Being named gets you a visit from someone who now knows one fact about you. The page they land on should answer the next question, not restate the first.

Leave the homepage until last. It is the page everyone wants to work on and the one least likely to be the page an engine quotes. What gets quoted is the page answering a specific buying question — a service, a price, an area, an availability. Those are the pages to write first, and the ones every listing about you should be made to match.

A worked example

From a measurement on a local market, business anonymised.

Baseline. Twelve buying questions, five repeats, four assistants. Named in 38% of general recommendation answers and 0% of price answers.

What the citations showed. For price questions, four sources came back again and again: two directories, a review platform, and a regional “best in the city” roundup. The business appeared on one of them, with an out-of-date service list and no prices anywhere.

What was done. Listings corrected and completed on all four, using the same service names and price ranges as the business’s own site. One page rewritten so the price question is answered in the first two sentences instead of in a table three screens down.

Result. General recommendation moved four points — reported as no change, because that is what four points is. Price answers went from zero to being named in roughly a third of them.

What we did not claim. That the listings caused it. We reported what was changed, when, and what moved after, and noted that the movement was concentrated in the question group the work targeted rather than spread evenly.

How this differs from the featured-snippet era

Anyone who optimised for featured snippets in 2018 will recognise most of the technique and none of the economics.

The practical shift: you are no longer competing for one slot against a ranking. You are competing to be one of several names, in an answer that changes every time it is generated, on a surface that may never send you a click. Optimisation still works; the reporting has to change.

Common questions from people who bought SEO before

“Is this just structured data?” No. Schema helps a machine parse your page correctly; it does not supply the reason to name you. See schema markup for AI search.

“Do I need llms.txt?” Probably not urgently. It is a proposal rather than a standard, and no major assistant has confirmed reading it — the honest version is here.

“Will my existing content rank in AI answers?” Some of it. Pages that answer a buying question directly tend to travel; pages that build to a conclusion tend not to.

A short glossary

Answer engine — anything that replies rather than lists: assistants, AI Overviews, voice assistants, featured snippets.

Extraction — the process of lifting a self-contained passage out of a page to answer a question. What most AEO technique is aimed at.

Self-contained passage — a section that answers its own heading without needing the paragraphs around it.

Share of answers — the percentage of answers naming a business across repeats, per engine. The metric that replaces rank.

Zero-click — an answer complete enough that nobody visits a source. Common, and the reason traffic stops being a usable measure of this work.

Cross-checking — a system comparing what several sources say about the same business. Why consistency matters more than polish.

Where to start

Measure once, before changing anything, so the next number has something to be compared to.

The free check asks the buying questions for your business and emails you what came back — no charge, no call. The AI search optimization guide covers the work in full, and how to improve AI visibility is the step-by-step version.

One thing to do tomorrow

Take your single most valuable buying question and check whether the page you would want an engine to quote answers it in its first two sentences. If it does not, fix that page before reading anything else about this subject. It is half an hour, it costs nothing, and it is the change most likely to be visible the next time anybody measures.