SEO gets you into search results. AEO, answer engine optimisation, is about being named when an assistant answers a question directly. GEO, generative engine optimisation, is about your content being used to build that answer, whether or not you get named.
That is the honest version of the distinction. The dishonest version is that these are three disciplines requiring three budgets. In practice AEO and GEO describe the same work from two angles, and both sit on top of SEO rather than replacing it.
Where the words came from
SEO you know. The other two are recent, and they did not arrive the same way.
GEO has an actual origin: a 2023 paper by Aggarwal and colleagues at Princeton and Georgia Tech, GEO: Generative Engine Optimization, later accepted to KDD 2024. It models a generative engine as retrieval plus synthesis and asks how content can raise its chance of being cited in the synthesised answer. It is a research term with a definition behind it.
AEO came from the industry, not from a paper. It grew out of the older idea of optimising for featured snippets and voice assistants, and got recycled once ChatGPT started answering questions directly. Nobody owns the definition, which is why you will see it used to mean slightly different things by every agency selling it.
So when someone insists on the distinction, ask which definition they are using. The answer tells you a lot.
What genuinely differs
| SEO | AEO | GEO | |
|---|---|---|---|
| You are optimising for | A ranked list of links | Being named in a direct answer | Being the material an answer is built from |
| Success looks like | Position on the results page | Your name appears, ideally first | Your page is cited or paraphrased |
| The unit of work | A keyword | A question | A fact worth quoting |
| How you measure it | Rank tracking, Search Console | Running the questions and reading the answers | Checking which sources the answer names |
The fourth row is where the real difference lives. SEO has fifteen years of measurement infrastructure behind it. The other two mostly do not, which is why so much AEO advice stops before telling you how to check whether it worked.
The part nobody puts in the pitch deck
Roughly eighty per cent of the work is identical across all three, and it is unglamorous.
Your company facts have to agree with each other everywhere they appear: your site, Companies House, your Google profile, every directory you are listed in. Your pages have to answer one question each, in the opening paragraph, with numbers you can defend. Your HTML has to be readable without running JavaScript. Other people have to say things about you somewhere a crawler can find. And search engines still have to be able to crawl and index you, because Perplexity and ChatGPT both lean on search results to decide who is even a candidate.
Do that and you are optimised for all three at once. Skip it and no amount of AEO tactics saves you, because there is nothing underneath for a model to trust.
Which one you need
If nobody can find you in Google, start with SEO. Not because the others do not matter, but because an assistant that cannot find you in a search index has no route to recommending you. This is most businesses, and it is a duller answer than they were hoping for.
If you rank fine and enquiries are still falling, look at the answer layer. When Google shows an AI Overview above your result, a share of the people who would have clicked never do. Ranking third on a page nobody scrolls is not the same win it was in 2022.
If you sell something people research by asking rather than searching, which increasingly means anything technical or expensive, then AI search optimisation stops being optional. Your buyers are asking a model to shortlist for them, and you are either on the list or you are not.
And if you are a local business whose customers find you on a map, do local SEO first and ignore the acronyms for now.
How to buy it without getting fleeced
This is a new enough market that pricing is untethered from substance. Four questions sort it out.
Ask what they will measure and how. If the answer is a screenshot of ChatGPT naming the client, that is an anecdote, not a measurement. Ask how often, across which models, and whether the wording of the prompts stays fixed between runs.
Ask what they would do first. If the honest answer for your business is basic SEO and consistent company data, a good supplier says so, even though it is a smaller invoice.
Ask whether the work survives a model update. Tactics that game a specific phrasing do not. Verifiable facts, consistent entity data and readable pages do.
And be careful with guarantees. Nobody controls what a model says about you. Anyone promising a placement in ChatGPT is either misunderstanding the mechanism or counting on you to.
Our own view, for what it is worth: we sell this work and we still tell more than half the people who ask about it to fix their site and their listings first. The order matters more than the label.
If you want the practical version rather than the taxonomy, our guide on how to get recommended by ChatGPT and AI search is the checklist behind all of this.