What AEO is, and why it matters more than it sounds like it should

Every few years the marketing industry produces an acronym that is mostly a repackaging exercise. AEO — answer engine optimisation — arrived looking like one of those. It is not, quite, and the reason is worth understanding before you decide whether to spend anything on it.

The underlying change is simple. A growing share of the research people do before buying now happens inside a system that answers rather than lists. Someone asks an assistant which managed IT provider handles healthcare compliance in their region, and gets a paragraph naming three companies. No results page, no ten blue links, no impression counted in your Search Console. If you were not one of the three, you were not considered, and nothing in your analytics stack will ever tell you it happened.

That is the whole thing. Everything else about AEO is a consequence of it.

Why the format of your content suddenly matters differently

Search engines have spent twenty years learning to understand messy pages. They are extremely good at it. A rambling introduction, a heading that says 'The Bottom Line' instead of asking a question, a statistic with no source — none of that stops a page from ranking if the page is broadly useful and the site is credible.

A language model assembling an answer is doing something different. It is looking for passages it can lift with confidence: statements that are self-contained, unambiguous about what they claim, and attributable. When it has to choose between a page that states the answer in its first sentence under a heading phrased as the question, and a page that gets there in paragraph four after some scene-setting, the first one is easier to use.

So the practical changes AEO asks for are small and slightly boring:

• Answer the question in the first sentence, then elaborate.
• Phrase headings as the questions people actually ask.
• Define your terms explicitly rather than assuming shared vocabulary.
• Attach facts to sources and dates instead of letting figures float.
• Put a short summary at the top that could stand alone as the answer.

None of these hurt search rankings. Most of them improve the reading experience for humans too, which is a hint that they were good practice already and the industry had drifted.

The part nobody wants to hear: your site is not the only input

Here is where AEO diverges from SEO in a way that is genuinely uncomfortable. When you optimise for search, your own pages are the primary lever. When an assistant answers a question like 'who are the best X in Y', it frequently synthesises from directories, review platforms, forum threads and roundup articles — sources you do not own.

This has two implications. The first is that consistency across those sources matters more than most businesses realise. If three directories describe you differently, and one of them has a category from 2019 that no longer reflects what you sell, that inconsistency shows up as absence rather than as a wrong answer. The second is that some of the most valuable AEO work is not content production at all. It is unglamorous listing hygiene.

We are saying this plainly because it cuts against our own interest. A platform that generates articles would prefer the answer to be 'publish more articles'. Publishing more articles helps. It is not sufficient.

How you know if it is working

The measurement problem is the reason AEO has been slow to get budget. There is no equivalent of a ranking position, because assistant answers are non-deterministic. Ask the same question twice and you can get different names. Ask it on a different day, after a model update, and the picture shifts again.

The workable approach is statistical rather than positional. Fix a set of prompts that represent how your buyers actually ask. Run them against each assistant on a schedule. Record whether you were named and which page, if any, was cited. What you get is a mention rate: named in six of ten runs on Claude, two of ten on ChatGPT, zero on Perplexity.

That is a number that moves in response to work, and it is defensible in a way that a single lucky answer is not. If your mention rate on a set of prompts goes from one in ten to five in ten over a quarter in which you published fifteen answer-format pages on those exact questions, you have evidence. Not proof — evidence.

Anyone selling you a fixed 'AI ranking' for your brand is describing something that does not exist. The output is probabilistic; the measurement has to be too.

The robots.txt question

Before any of this, check whether you have accidentally opted out. A surprising number of small business sites picked up blanket AI crawler blocks in 2023 and 2024 from a plugin default or a template someone copied, and nobody has looked since.

The agents to look for are GPTBot, OAI-SearchBot, ClaudeBot, PerplexityBot, Google-Extended and CCBot. If your robots.txt disallows them, you have removed yourself from the retrieval pool for every assistant that honours the file — which is the ones worth appearing in.

There are legitimate reasons to block. If your content is your product and you are pursuing licensing, blocking is a negotiating position. If your website exists to win customers, blocking is working against the purpose of the website.

What this is worth right now

We would rather be accurate than exciting about the size of this. For most small businesses today, classic search still sends the large majority of discovery traffic. AEO is not where your budget should move wholesale.

What makes it worth doing now is the cost of doing it. The formatting changes are nearly free if you are producing content anyway. The listing consistency work needed doing regardless. The measurement takes a scheduled job. Against that, the downside of ignoring it is being absent from an increasingly common research step with no way to know you are absent.

That asymmetry — low cost, invisible downside — is the argument. Not a revolution. A cheap hedge on a channel that is clearly growing, plus a set of writing habits that make your content better for everyone reading it.

Where to start

1. Read your robots.txt and decide about AI crawlers deliberately rather than by default.
2. Write down the ten questions a buyer asks before choosing someone in your category.
3. Run those questions against ChatGPT, Claude, Gemini and Perplexity. Write down who gets named.
4. Pick the three where you are absent and where you have a genuine answer, and write those pages properly — answer-first, defined, sourced.
5. Re-run the prompts monthly and watch the mention rate rather than any single answer.

That is a week of work spread over a quarter, and at the end of it you will know something about your visibility that most of your competitors do not know about theirs.