AEO vs SEO — what actually changes, and what does not
AEO is sold as a replacement for SEO. Most of it is the same work marked more strictly — here is the part that is genuinely different.
Two acronyms are being sold as separate products, usually by the same person, usually on the same invoice. It is worth knowing which parts of the second one are real.
Answer engine optimisation is search engine optimisation marked more strictly. A search engine ranks documents and lets the reader choose; an answer engine reads documents and chooses for them. Almost everything a page needed in order to rank well, it still needs — the difference is that a human reader forgives a page that buries its answer, and a model simply skips it and quotes a competitor.
What does not change
Speed, crawlability, one clear subject per page, internal links that make the hierarchy obvious, and content somebody actually wants. Every one of these is a thirty-year-old requirement and every one still applies. Answer engines are largely reading pages that search engines already found and understood. A site that fails at search usually fails at being quoted for identical reasons.
What genuinely changes
Extraction replaces ranking. In search, a page competes as a whole document for a position. In an answer, one passage competes to be the thing that gets lifted. That makes the internal structure of a page matter in a way it did not before: a heading that matches the question, and directly under it a paragraph that answers it completely and survives being quoted with nothing around it.
Consistency becomes a ranking-scale problem. A generated answer is assembled from several sources at once. If your own site, your Google Business Profile and three directories disagree about what you do or where you do it, that disagreement is not a small untidiness — it is the reason the model names someone else.
Being described off your own site starts to matter more. A model reconciling sources needs sources to reconcile. A business that exists only on its own website gives it nothing to be confident with.
What is being oversold
"AI schema." There is no such thing. No structured data type confers eligibility to be cited by an assistant, and Google's own generative-AI guidance names none. Structured data is still worth shipping because it states plainly what a page is — but it is not a lever on citation, and anybody selling it as one is selling a promise they know is empty.
llms.txt. A reasonable convention, cheap to publish, and adopted by no major model provider as a commitment. This site ships one because it costs nothing and can be audited in a click. It should not appear on an invoice.
GEO as a second product. Generative engine optimisation and answer engine optimisation were coined by different people for overlapping work. If someone is billing for both, the fair question is what the second one does that the first does not.
The short version
If your SEO is genuinely good — fast pages, clear structure, honest content, consistent facts — you are most of the way to being quotable already. The remaining work is a rewrite of how answers sit on the page, and it is a smaller job than the acronym implies. If your SEO is poor, no amount of AEO will help, because the pages an assistant reads are the pages a search engine already declined to rank.