What is AEO? Answer Engine Optimisation, explained
· 7 min read · ActiScore
Answer Engine Optimisation is the work of getting recommended when someone asks an AI assistant who to hire, rather than getting ranked when someone types a query into a search box. It is a different question, measured a different way, and most businesses have never checked their answer.
What actually changed
Search used to end with a list. You typed something, you got ten links, and the work of ranking was the work of getting found. A person then chose among the results — which meant being fourth still put you in the running.
An assistant does not return a list. Asked “who should I call for managed IT in Columbus?” it returns a short answer naming three or four companies. There is no fourth page. There is barely a second screen. You are named, or the conversation happens without you.
Ranking is a position among many. Being named is binary. That difference is the whole reason AEO is measured separately from SEO.
AEO and SEO are not the same job
They overlap, and good SEO helps. But they answer different questions and fail in different ways — a site can be technically excellent and still never be named.
| SEO | AEO | |
|---|---|---|
| The question | Where do we rank for a term? | Are we named when someone asks? |
| The result | A position, 1 to 100 | Named or not named |
| Who decides | A ranking algorithm | A model summarising sources |
| How you check | Rank tracking | Ask the assistants and read the answer |
| Failure mode | Ranking below a rival | Being absent from the answer entirely |
The practical consequence: rank tracking cannot tell you whether you have an AEO problem. A business ranking third for its main term can be named by none of the four major assistants, and nothing in a rank report would show it.
How to actually measure it
There is only one honest method, and it is the obvious one: ask the assistants the questions a customer would ask, and read what comes back. Everything else is inference.
- Ask buying-intent questions, not brand questions. “What do you know about Acme Plumbing?” tests recall. “Who are the best plumbers in Denver?” tests whether you get recommended, which is the thing worth money.
- Ask each assistant separately. Being named by ChatGPT and missed by Gemini is a different problem from being missed by all of them, and a single blended score hides which one you have.
- Ask more than once, in more than one phrasing. One question is an anecdote. Five phrasings of the same intent is a measurement.
- Record who was named instead. The companies that appear in your place are the most useful output of the whole exercise.
Being named is not the only thing to check
An assistant can recommend you and still describe you wrongly — the wrong phone number, the wrong city, services you dropped two years ago. That is worse than being absent, because the customer acts on it. A complete check answers four questions:
- Awareness — can the assistant find the business at all?
- Accuracy — are the details it gives correct?
- Reputation — how positively does it describe the business?
- Recommendation — does it actually suggest you to a customer?
What actually moves the needle
Assistants answer from two places: what a model absorbed during training, and what it can retrieve right now. You cannot influence the first directly. You can influence the second, and you can influence what the next training run absorbs.
- Be unambiguous about who and where you are. Structured data — Organization or LocalBusiness markup with a name, phone and locality — is the one place a business states its own facts in a form with no interpretation required. It is also what an accuracy check compares against.
- Answer real questions in plain prose. A page that answers “how much does managed IT cost in Ohio?” directly is quotable. A page of adjectives is not.
- Be consistent everywhere else. Assistants that retrieve live sources read directories, review sites and profiles. A phone number that differs across three of them is a fact the model has no way to resolve.
- Be present in the sources that get cited. Retrieval-based assistants cite what they can reach. Being absent from the places that get quoted is being absent from the answer.
- Keep the site fast and crawlable. Unremarkable, and still true: a page that cannot be fetched cannot be quoted.
The uncomfortable part: some of this takes months to show up, because a model that answers from training data will not know about your change until it is retrained. Retrieval-based assistants move faster, which is why they are the ones worth re-measuring monthly.
Where to start
Measure before you change anything. You need a baseline, and more importantly you need to know which of the four questions you actually fail — the fixes are different.
You can run that check on your own site free, with no signup, on our AI visibility checker. It asks each assistant the questions a customer would, shows you the answers verbatim, and names the companies recommended in your place.
Common questions
What is AEO?
Answer Engine Optimisation is the work of getting named when someone asks an AI assistant like ChatGPT, Gemini, Claude or Perplexity who to hire, rather than getting ranked in a list of search results. Where SEO produces a position among many, AEO is binary: you are named in the answer or you are not.
How is AEO different from SEO?
SEO asks where you rank for a term and returns a position from 1 to 100. AEO asks whether an assistant names you when a customer asks who to hire, and the answer is yes or no. They overlap, but a site can rank well and still be named by none of the major assistants, and rank tracking will not show it.
How do you measure AI visibility?
Ask each assistant the buying-intent questions a customer would ask, in several phrasings, and read what comes back. Record whether the business was named, whether the details given were correct, how positively it was described, and which companies were recommended instead. Ask each assistant separately, because being named by one and missed by another is a different problem from being missed by all.
Can you influence what AI assistants say about your business?
Partly. Assistants answer from training data, which you cannot change directly, and from live retrieval, which you can influence. Publishing unambiguous structured data, answering real questions in plain prose, keeping your details consistent across directories and review sites, and being present in sources that get cited all help. Retrieval-based assistants reflect changes fastest.
How often should AI visibility be re-checked?
Monthly is enough for most businesses. Assistants that answer from live search can change their recommendations as sources change; assistants that answer from training data will not move until they are retrained. Re-checking monthly catches the first kind and builds a record of the second.