How AEO works
AEO stands for answer engine optimisation: the work of making a page the thing an answer engine reaches for when it writes a reply — Google’s AI Mode and AI Overviews, ChatGPT, Perplexity and the rest — and, ideally, names as its source.
The mechanism it aims at is retrieval rather than ranking. These systems break a question into parts, pull passages that answer each part, and assemble a reply from them. So the unit that competes is a passage, not a whole page: a short, self-contained, unambiguous statement that would still make sense if it were lifted out and pasted somewhere else. Anything that only makes sense in the context of the paragraphs around it is hard to quote, and tends not to be.
Why AEO matters
Because a growing share of questions now end without a click. When that happens, the only trace of you left in the exchange is whether you were named and whether what was said about you was accurate. A business quoted correctly inside an answer is present in the decision even when the visit never happens.
It also rewards clarity that helps human readers anyway. Writing the answer before the background, using the customer’s words, and giving each page one job are old editing virtues. The difference now is that a machine punishes the alternative more visibly than a bored reader ever did.
Where AEO goes wrong
The most common failure is cosmetic: pages sprouting question-shaped headings with no real answer beneath them, or a long FAQ block repeating what the page already said. Adding the shape of an answer is not the same as adding an answer, and it makes the page harder to read.
The second is the promise problem. There is no submission process, no paid slot and no guarantee inside an organic AI answer, so any offer of a guaranteed citation is selling something the seller does not control. The third is abandonment — dropping ordinary search work because AI answers feel newer. Most enquiries still arrive through links.
What to do about it
Start with the questions you are actually asked by phone, email and WhatsApp, and give the important ones a page each, with the answer in the opening lines and the caveats after it. Use plain markup so a parser can tell structure from decoration, keep your prices, service names and coverage identical wherever they appear, and make sure the crawlers feeding these systems are not blocked from the pages you want quoted.
Then judge it on evidence rather than feeling. Put your customers’ questions to the engines yourself and record whether you are named and correctly described, watch impressions against clicks in Search Console, and ask new enquiries what they had already been told. The related discipline aimed at longer generated summaries is generative engine optimisation, and the site’s answer engine optimisation page sets the same ideas out in a service context.
