The measurement is free to start: one field, your domain, 8 questions across all four engines. A full audit is a one-off fee, and the monthly programme is rolling and can be cancelled at any time. Every price is on the pricing page — nothing is quoted only on a call.

The category, explained

AEO, GEO, LLM SEO. Four names, one question.

Q: What is an AEO agency?

An AEO agency is a supplier that works on whether AI assistants name a business in their answers — measuring how often it happens today, diagnosing why it does not, and doing the technical, editorial and third-party work that changes it. The category is about two years old, so the label is worn by everyone from research-led specialists to SEO agencies who renamed a service line. The distinction that matters is not the label but whether the supplier measures before and after, and hands you the raw answers.

Q: Is GEO the same thing as AEO?

In practice, yes. GEO (generative engine optimisation) and AEO (answer engine optimisation) describe the same work: getting named in the answer an AI assistant gives, rather than in a list of links. GEO is named after the generative model, AEO after the answer. Some people draw a line — AEO for direct-answer formats like featured snippets and voice, GEO for large language models — but no supplier in this market uses the terms consistently enough for that distinction to help you buy.

Q: Is AEO just SEO with a new name?

No, though it overlaps and good SEO helps. Three things differ. The unit is a named mention rather than a ranked link, so you either appear or you do not — there is no position four. The answer is assembled from several sources at once, so what a directory or a review site says about you can beat your own website. And the output is not stable: ask the same question twice and the list can change, which makes repeat measurement part of the method rather than an optional extra.

Q: Can anyone guarantee that ChatGPT will recommend my business?

No. Nobody controls what these engines say, the models are retrained and the retrieval layers change without notice, and the same question can produce different names on consecutive runs. A guaranteed mention is either a guess or something being gamed in a way that will be patched. What can honestly be promised is a measurement you can audit, a diagnosis of each gap with its evidence, and work that is checked against the next measurement.

Q: How long does it take to show up in AI answers?

The technical layer — crawlability, schema, answer formatting, pages that match real questions — usually shows up within one or two monthly runs, because it changes what the engine can read and extract immediately. The source layer takes longer: getting into the directories, review sites and publications the engines actually read for your category is a matter of months, and it is usually where the larger and more durable gain sits.

Q: Do AI assistants matter yet for a business my size?

It depends on whether your buyers ask questions before they choose, and whether you can tell. For a considered purchase — a clinic, a law firm, a supplier chosen after research — the assistant is increasingly the first shortlist, and the shortlist is three to five names. The honest way to find out is to measure rather than argue about forecasts: run the free audit, see whether you are named on the questions that matter to you, and decide with a number in front of you.

Answer engine optimisation is the work of getting a business named when someone asks ChatGPT, Gemini, Claude or Perplexity for a recommendation. This page is what we would want to read before buying it: what the terms mean, how the engines actually decide, what can and cannot be changed, and what to ask a supplier — including us.

Measure your domain free See measurements we published

Terminology

The four terms mean the same thing.

This is a young category and nobody has won the naming argument. If a supplier tells you their term is the real one and the others are outdated, that is positioning, not a distinction.

Answer engine optimisation

AEO

Getting named in the answer itself, rather than in a list of links below it. The term predates the current AI assistants — it started with featured snippets and voice — but it is now used mostly for ChatGPT, Gemini, Claude and Perplexity.

Generative engine optimisation

GEO

The same work, named after the generative model rather than the answer. The term comes from a 2023 academic paper and is the one most used by people who came to this from a technical background.

Large language model SEO

LLM SEO

The same work again, named to reassure people who already buy SEO. Useful as a search term, misleading as a description: ranking and being named are related but they are not the same measurement.

AI visibility

The measurable outcome the other three are aiming at: how often, and how accurately, an engine names you when a buyer asks a question you should win. This is the one we use, because it names a number rather than an activity.

We use AI visibility for the outcome and treat the rest as search terms.

How the decision is made

An engine names you in three stages, and each one can drop you.

Most advice in this field treats the engine as a black box and then sells confidence about it. It is not a black box: the stages are observable, and which one you are failing at is the single most useful thing to know, because the fix is completely different in each.

It decides what to read

The engine rewrites the buyer’s question into one or more searches, runs them, and takes a handful of results — often five to fifteen pages. Nothing outside that set can appear in the answer, however good it is. If your pages are unreadable to the crawlers, or no page of yours matches how the question was rewritten, the work stops here.

Levers: Crawlability · a page per real question · being present in the sources it searches

It decides what it can extract

From those pages, the model lifts what it can resolve in a single pass: a direct opening answer, a list, a table, marked-up facts. A price in paragraph nine, a claim inside an image, a service buried in a PDF — these are read as absent. This is why two businesses with the same substance get different answers.

Levers: Answer-first writing · tables and lists · schema · one fact in one place

It decides who to name, and resolves conflicts

The model then composes a recommendation, usually naming three to five businesses, and where sources disagree it picks one version. Directory entries, review sites and press carry real weight here, and an out-of-date third-party fact can beat your own site. The names it chooses vary between runs, which is why a single run tells you almost nothing.

Levers: Third-party sources · consistency of facts across the web · repeat measurement

Being named is not one thing you are good or bad at. It is three, and a business can be excellent at stage two while failing stage one outright — which is why a plan that starts with content, before anyone has checked whether the crawlers can read the site, wastes the first two months.

The three stages in more depth → · How we measure →

The part nobody sells

In a regulated sector, being named wrongly costs more than not being named.

Every supplier in this category sells appearances. Almost none of them measures what the engine actually says about you once you appear. We check every claim against your own site: confirmed, contradicted, or unverifiable.

How citation fidelity is measured →

Choosing a supplier

Five questions to ask anyone selling this — us included.

What is my number today, how many times do you ask each question, what can I check myself, how will we know it worked, and what do you refuse to do. Each with our own answer next to it.

Read the five questions →

Questions

The ones people actually ask.

What is an AEO agency?

Is GEO the same thing as AEO?

Is AEO just SEO with a new name?

Can anyone guarantee that ChatGPT will recommend my business?

How long does it take to show up in AI answers?

Do AI assistants matter yet for a business my size?

What does it cost?

Something here you want argued properly rather than in a paragraph? Write to hello@namedfirst.co.uk.

The argument is cheap. The measurement is free.

One field, your domain, 8 buyer questions across all four engines. If they already name you everywhere, the report will say so and you should not buy anything.

Run the free audit