Mostly it’s a new label, but not entirely. Most of the work that gets your business named by ChatGPT or quoted in an AI Overview is work that was already good SEO, and Google said as much in its own guidance in May 2026: optimising for AI search is still SEO. There are five places where the two come apart, though, and they change what you measure and where the effort goes.
If you want the background on what generative engine optimisation is in the first place, that’s our guide to GEO. This piece is for the more sceptical question underneath it. Is someone selling you something new here, or is it the same job with a fresh acronym on the invoice?
Start with the part that hasn’t changed
The unglamorous answer is that most of it hasn’t, and that’s the important thing to say first.
An AI tool answering a question in real time is reading the same web as everyone else. If a page can’t be crawled it won’t be in anyone’s index, and a page that isn’t in an index can’t be found and quoted. We’ve seen recruitment sites where the job listings are built by a script that search engines never see the output of, so those vacancies are invisible in Google and equally invisible to ChatGPT. It’s the same fault with the same fix, and there’s no new discipline involved in spotting it.
Answering a question clearly and early isn’t new either. Google’s featured snippets (the boxed answer that sometimes sits above the normal results) rewarded exactly that behaviour for the best part of a decade. Anyone who was writing for those is most of the way to writing for an AI answer already.
The timescale hasn’t changed either. Building the kind of authority these systems reward takes months, in the same way ranking does, so GEO isn’t a shortcut for a site that hasn’t earned anything yet.
Where the two come apart
Five things, and each one changes something practical about how you work.
1. You’re competing for a mention rather than a position
In classic search there are ten places on the first page and you either hold one of them or you don’t. An AI answer has no ordered list and no position three to defend, just a paragraph that names two or three sources, and the only question is whether you’re one of them. That sounds like a technicality until you notice what follows from it: several competitors can win the same question at once, and you can be present in an answer without being the top-ranked page for anything.
2. The thing being quoted is a passage, not the whole page
AI tools usually break a question into smaller searches and then combine what comes back, so if you ask one for help specifying a fastener for a marine environment it might search separately for corrosion resistance, for material grades and for supplier lead times. The page it quotes for the middle part could be a technical note three levels down on an engineering firm’s site, on a page that ranks nowhere for anything. In classic SEO you optimise a page for a query, but here a single well-written passage can do the work on its own.
3. What other people say about you counts for more than it used to
Links have always mattered, but AI tools lean heavily on third-party descriptions – trade press, directories, forums, review sites – when they’re working out what your business is and whether to name it. An unlinked mention in a sector publication can carry weight that a link-building campaign wouldn’t, which pulls the work away from your own site and towards PR, and that isn’t where most SEO budgets sit.
4. You can be doing well and see nothing in your analytics
If someone reads an answer that names you and doesn’t click, there’s no referral, no session and no line in your reporting. What you might see instead, a week later, is your business name typed into Google, which is the same person arriving by a route your reporting files under branded search. Measuring this properly means accepting that some of the effect is inferred rather than counted, which is uncomfortable if you report to a board that likes a number.
5. There’s no position to hold, because the answers keep changing
Ask the same AI assistant the same question twice and you can get two different replies citing two different sets of sources. Nothing here behaves like holding fourth place for six months, so progress is harder to demonstrate, and anyone promising you a guaranteed ranking in ChatGPT is promising something the tool doesn’t offer.
So why does the label persist?
Partly for an honest reason. The term came out of a 2023 academic paper rather than a marketing department, and the researchers had a real distinction in mind: optimising for a system that writes an answer, rather than for one that returns a list of links. That’s a fair thing to give a name to.
And partly for a less honest one. “Generative engine optimisation” is easier to put on a proposal than “the SEO you should have been doing since 2019”, because a new acronym creates a new budget line, and a new budget line is worth a lot to anyone selling a tool. That’s why AEO, AIO and LLMO are all representing the same idea, rather than because five separate disciplines appeared at once.
Is it worth paying someone for GEO, then?
The useful test is to ask what’s in it.
If the deliverables are a schema audit, an llms.txt file (a plain text file some now recommend adding to describe your site to AI tools) and a monthly report on whether you appear in ChatGPT, you’re being sold a checklist. Schema markup is worth having for other reasons, and llms.txt is barely used by the tools that do the citing, so neither of them is the thing that gets you named.
If the deliverables are content written by someone who knows your field, a site clean enough to be read, and a serious effort to get you mentioned on the sites your buyers trust, then it’s worth paying for. That’s also SEO, done with the emphasis in the right place. What matters is the list of work underneath the name, rather than the name itself.
The one thing we’d treat as a warning sign is anyone offering GEO with no SEO underneath it, because a site that can’t be crawled or has nothing worth quoting won’t be rescued by optimising for AI tools. The AI tool is reading the same broken site.
Where we’ve landed
We sell both, and we sell them as one thing, because for almost every business we work with they are one project. We could package GEO separately and charge for it, and we’d rather explain why we don’t.
What would change our minds is evidence that AI tools are weighing something classic search doesn’t touch at all: a signal you could only earn by working on it directly, with no SEO benefit attached. There are claims to that effect, and so far, when we’ve followed them back to the testing behind them, the evidence has been thin. The field is only three years old, though, so the honest answer now might not be the honest answer in a year. We’ll keep checking, and we’ll say so here if it changes.
If you want a straight read on where your site stands, without the acronym, our SEO and GEO service page sets out how we work.
