Answer Engine Optimisation (AEO) is the practice of making your website easy for AI assistants to retrieve, understand and cite. Where SEO aims to win a click from a results page, AEO aims to be the source an answer is built from — inside ChatGPT, Perplexity, Claude or Google’s AI Overviews, where there may be no results page at all.
Why this is suddenly urgent
A growing share of commercial research now ends inside an assistant rather than on a search results page. Someone asking “who builds websites for small businesses in London” increasingly gets a synthesised answer naming three or four companies. If you are not one of the cited sources, you are not in the consideration set, and no amount of ranking fourth on Google will put you there.
The uncomfortable part is that being cited is not the same as ranking. Assistants pull from pages they can fetch quickly, parse cleanly, and attribute confidently to a real organisation. Those are three separate problems.
The three things that actually work
1. Let the retrieval bots in
This is the highest-leverage and most commonly botched step. AI assistants use two distinct classes of crawler, and they need to be treated differently:
- Retrieval bots fetch a page at the moment a user asks a question.
OAI-SearchBot,PerplexityBot,Claude-SearchBot,ChatGPT-UserandPerplexity-Userare the main ones. Blocking these removes you from AI answers immediately and completely. - Training bots collect data for future model versions.
GPTBot,ClaudeBot,CCBotandGoogle-Extendedsit here.
Plenty of sites block all of them by copying a robots.txt snippet from a publisher who had good reasons to. For a marketing site those reasons do not apply — you want to be found. Check your own file first; this is often a one-line fix worth more than a quarter of content work.
2. Answer the question in the first paragraph
Assistants extract passages, not pages. A page that opens with three paragraphs of throat-clearing before reaching the point gives a model nothing clean to quote. A page that opens with a direct 40–60 word answer, then expands, gives it exactly what it needs.
The same structure helps human readers, which is why this is the rare optimisation with no downside. Put the answer first, then the caveats, then the detail.
3. Make your entity unambiguous
A model has to be confident that “Websify” the website builder is one specific company before it will name you. That confidence comes from consistent signals: the same organisation name, address and registration number everywhere, a stable Organization identifier in your structured data, and sameAs links to the profiles that corroborate it.
This matters doubly if you operate across markets. We publish in the UK and Hong Kong from a single organisation identity rather than two look-alike entities, precisely so the signal is not split in half.
What is overrated
Two things get more attention than their results justify.
Piling on schema types. Structured data helps a machine parse your page, and FAQPage, Service and BlogPosting markup are all worth having. But there is no evidence that adding a tenth schema type lifts citation rates. Schema is table stakes, not a lever.
llms.txt. The proposed convention for giving models a curated index of your site is cheap to add and almost never fetched. We publish one because it costs nothing. Do not build a strategy on it.
Where to start
If you only do one thing this month, open yourdomain.com/robots.txt and confirm the retrieval bots are allowed. Then rewrite the opening paragraph of your five most commercially important pages so each one answers its question directly.
If you would rather not do any of it yourself, that is roughly what our monthly plans cover — the technical layer is handled and the content follows a fixed structure. You can also see how the process works before committing to anything.