GEO and AEO
Generative engine optimisation, minus the part where we pretend it is new.
Most of what is sold as generative engine optimisation is technical SEO with a different noun on the invoice. Three things genuinely differ: models reward answers stated plainly rather than built up to, attribution depends on your entity being described consistently across the web, and measurement is a real problem because there is no rank to track. This page is about those three, and about what we will not claim.
Straight answer
Most GEO services are selling you SEO you should already have.
That is not a reason to ignore it. It is a reason to know which parts are genuinely new, so you are not paying a premium for work that was already in scope.
Sold as new, actually just SEO
- Structured data that validates
- A site a crawler can read without fighting it
- Named authors with real expertise
- Clear headings and a logical page structure
- Fast, stable pages
- Accurate, current facts on the page
Genuinely different
- Answers stated in self contained passages a model can lift
- Entity consistency across the sources models already trust
- Being described correctly, not merely being present
- Measurement by prompt sampling rather than rank tracking
- Third party pages about you mattering more than your own
- No submission, no console, no appeal process
The claim we will not make
Nobody can guarantee a citation in ChatGPT or an AI Overview. Outputs are non deterministic, they change between model versions, and the same prompt can return different sources on consecutive runs. There is no submission process and no support channel. What can be raised is the probability that you are found, extracted and described correctly. An agency promising more than that is describing something it does not control.
Evidence
We shipped the artefacts, and you can fetch them now.
A great deal of GEO is sold by people who have never implemented any of it. Everything below is live on this domain and verifiable in your browser in about ten seconds.
llms.txt
A machine readable index of every page on this site with a description of each. Generated from the same catalogue that builds the sitemaps, so it cannot drift out of date. Fetch it.
llms-full.txt
The long form version, carrying substantive content per page rather than a line each, for systems that want the body rather than the map. Fetch it.
A validated schema graph
Every page on this site carries JSON-LD that parses and validates, checked on each build rather than assumed. Entities are connected rather than emitted as isolated blocks.
Said plainly, because it matters here
There is no public evidence that any major AI provider reads llms.txt, and Google has said outright that it does not. We publish it because it costs almost nothing and might one day be useful, not because it is the mechanism by which anyone gets cited. If an agency is selling llms.txt as the core of a GEO package, that tells you how closely they have followed this.
Scope
What generative engine optimisation covers here.
Weighted towards the parts that are genuinely different, since the rest is technical SEO and is priced as such.
Extractable answers
Pages restructured so the answer sits in a self contained passage a model can lift without the surrounding paragraph. Most commercial copy buries the answer under a build up.
Entity clarity
One unambiguous description of who you are, what you do and where, consistent across your site, your profiles and the third party sources that describe you.
The sources models lean on
Assistants disproportionately cite a small set of reference sites and communities. Being described accurately there frequently matters more than anything on your own domain.
Structured data that connects
A graph where organisation, people, services and articles reference each other, rather than isolated blocks a parser cannot relate to one another.
Prompt sampling
A defined set of real customer questions run repeatedly across the major assistants, recording presence, description and co-citation. Coarse, and reported as coarse.
Correcting the record
Being described wrongly is more common than being absent and does more damage. Finding where the wrong description originates and fixing it at the source.
GEO, AEO and SEO
The three terms, and whether the difference matters.
SEO is getting a page to rank in a list of results. It is still where almost all the traffic and almost all the revenue is, for nearly every business.
AEO, answer engine optimisation, is getting your answer used when a system responds directly instead of listing links. In practice this means featured snippets and AI Overviews as much as it means chat assistants.
GEO, generative engine optimisation, is the same objective aimed at generative systems specifically. The two terms are used interchangeably by nearly everyone, including the people selling them, and the distinction is not worth a meeting.
What is worth your attention is that all three rest on the same foundation. A site that cannot be crawled cleanly, whose facts are stale, whose authors are anonymous and whose structured data is broken will fail at all three simultaneously. That is why we would rather fix the foundation than sell you a separate GEO retainer running alongside your SEO one.
Cost
What it costs.
The same £95 blended hourly rate as everything else. There is no premium for the word generative.
Visibility baseline
From £650
A defined prompt set run across the major assistants, recording where you appear, how you are described and who is cited alongside you, with the inaccuracies traced to source.
Extraction and entity work
Scoped
Pages restructured so the answers are liftable, the entity graph connected, and the descriptions made consistent across the sources that carry weight.
Retained, 8 hours a month
£760/month
Sampling repeated on a cycle so you have a trend rather than a snapshot, with the underlying technical and content work running alongside it.
If your site has unresolved indexation or structured data problems, that work comes first and we will say so, because it serves both surfaces.
FAQ
Questions clients actually ask
What is generative engine optimisation?
Generative engine optimisation, or GEO, is making a site legible to systems that answer a question directly rather than returning ten links: ChatGPT, Gemini, Perplexity, Copilot and Google's AI Overviews. Answer engine optimisation, or AEO, is the same idea described from the other end. The terms are used interchangeably and arguing about the distinction is not worth your time.
You will also see it written as generative engine optimization, and answer engine optimization, in American English. Same thing. This site uses British spelling throughout and both forms refer to identical work, so if you arrived here searching either one you are in the right place.
The useful definition is narrow. It is not a new discipline that replaces SEO. It is the subset of technical and content work that decides whether a model can extract a clear, attributable claim from your page, plus a genuinely new measurement problem, because you can no longer see your own ranking.
Is GEO actually different from SEO, or is it repackaged?
Mostly repackaged, and anyone telling you otherwise is selling. The things that make a page citable by a model are largely the things that made it rank: a clear answer near the top, accurate structured data, real expertise attached to a named author, and a site a crawler can read without fighting it.
Three things do genuinely differ. Extraction rewards a page that states its answer plainly in a self contained passage, rather than building to a conclusion. Attribution depends on your brand being described consistently enough across the web that a model associates the claim with you. And measurement is a different problem entirely, because there is no rank tracker for an answer that is generated fresh each time.
Does llms.txt actually do anything?
Honestly, there is no public evidence that any major AI system reads it. Google has said publicly that it does not use it. No large provider has committed to supporting it. Anyone presenting llms.txt as the mechanism by which you get cited is overstating a proposal that has not been adopted.
We publish one on this site anyway, along with a full text version, because the cost is close to zero and the downside is nothing. It is a bet with a tiny stake, not a strategy. You can fetch llms.txt and llms-full.txt right now and see exactly what we generate, which is more than most agencies selling GEO can show you.
Can you guarantee we get cited by ChatGPT?
No, and neither can anyone else. Model outputs are non deterministic, they change between versions, and the same question asked twice can produce different sources. There is no submission process, no ranking to climb, and no support channel.
What can be done is raising the probability: making the claims extractable, making the entity unambiguous, and making sure the sources these systems already lean on describe you accurately. Anyone offering a guarantee is describing something they cannot control.
How do you measure whether any of this worked?
By sampling rather than tracking. A defined set of prompts a real customer would ask, run repeatedly across the major assistants, recording whether you appear, how you are described and what gets cited alongside you. Repeated over time it produces a trend, which is the honest ceiling of what this measurement can be.
It is coarser than rank tracking and it should be presented that way. What it does catch reliably is being described wrongly, which is more common than being absent and considerably more damaging.
Should we do this instead of SEO?
No. For almost every business the traffic and the revenue are still overwhelmingly in classic search, and a site that cannot be crawled properly will not be cited by a model either, because the same infrastructure underlies both.
Treat it as a second surface that the same foundational work serves, with a modest amount of additional effort on extraction, entity clarity and measurement. If someone proposes replacing your SEO budget with a GEO retainer, that is a reason to be suspicious of them.
Keep reading
Related services
Book me
Want to know how you are described right now?
Send your domain and the questions a customer would actually ask. We will run them across the major assistants and tell you where you appear, how you are described, and whether any of it is wrong.