Research · Agencies
Does AI name AI visibility agencies?
We measured 81 agencies that sell AI visibility, answer engine optimisation or generative engine optimisation — using the question their own prospective clients would type. Most of them are not in the answer.
The finding
agencies appeared in none of the three questions we ran in their own category and country.
That is 75% of the agencies we measured, with a margin of ±10.9 percentage points at 95% confidence. The median appearance rate across the whole set is 0%.
The cut that matters most is narrower. Of the 67 agencies whose own positioning is explicitly answer engine optimisation or generative engine optimisation — the ones selling this exact outcome — 53 appear in none of the answers.
In the United Kingdom, where we measured 57 of the 81, the figure is 48 of 57.
The part that surprised us
had their own website read as a source in their own measurement.
Not one. The assistants answered these questions by reading Google and a handful of other agencies' pages — never the page of the agency the question was about.
This is the part that changes what you do on Monday. A site can be perfectly readable and still not be what the assistant reads when it decides who to name. Those are two different problems, and only the first one is fixed by touching your own website.
What the assistants read instead
An assistant that names no agency is not answering from nowhere. Across the 81 measurements, the same handful of sources keeps coming back.
| Source | In how many measurements | How an agency gets in |
|---|---|---|
| google.com | 68 of 81 | Search results, not a page you own |
| clickslice.co.uk | 11 of 81 | Another agency published the comparison |
| polarisagency.com | 11 of 81 | Another agency published the comparison |
| agencies.semrush.com | 7 of 81 | A directory — you can list yourself |
| clutch.co | 6 of 81 | A directory — you can list yourself |
| ranktix.co.uk | 6 of 81 | Another agency published the comparison |
| sortlist.com | 5 of 81 | A directory — you can list yourself |
Two kinds of page carry almost all of it, and they are not equally open. The directories — Semrush's agency listing, Clutch, Sortlist — have no gatekeeper: you submit, you are in. The rest are other agencies' articles, which means a competitor wrote the page that the assistant reads to decide who is worth naming.
The practical reading is that the open route is open today and nobody is queueing for it.
Who does get named
Twenty of the 81 appear in at least one answer. These are the highest, and we name them because appearing is the good outcome:
| Agency | Appearance rate |
|---|---|
| graphek.de | 100% |
| margen.net | 67% |
| avantgrade.com | 50% |
| netgrade.de | 50% |
| clickslice.co.uk | 33% |
| fattoretto.agency | 33% |
These are rates over the runs we made, and the number of runs behind each one is small — small enough that 100% is one good day as easily as a durable position. What this set supports is the proportion of agencies that appear at all. It does not support ranking one against another, and we are not publishing it as one.
Why we measured this at all
Our first edition measured 64 online shops and found 48 of them in none of the answers. The obvious next question was whether the people selling the fix were in a different position. They are not.
That is not an accusation. An agency sells the work; whether the work has landed yet is a separate question, and the honest answer for most of this category is that it has not landed for themselves either. The reason to publish it is that a number nobody can check is a sales claim, and a number anybody can check is a starting point.
Method
For each agency we generate three buying questions from its own stated category and country — the kind a prospective client types, like "best AEO agency companies in the UK" — and run them through an assistant with live web search. Appearing means the agency was named or linked in the answer. Models are not deterministic, so every figure is a sampled rate over repeated runs, never a yes or no from a single pass.
This is not a random sample of agencies. It is the agencies our own collection found — from the WordPress.org plugin directory, from agency directories, and from names the assistants themselves cited in earlier measurements. Who is reachable by those routes is not who exists, and a different collection would produce a different set. Three questions per agency is also a coarse instrument: it supports the proportion appearing at all, and does not support ranking one agency against another. Individual agencies are not named in the negative; the dataset below is published without domains.
The dataset carries country, category, questions asked, appearance rate, whether the agency was named at all, and how many sources the assistant read. No domains, no URLs.
How to cite: Goppa Research (2026). Does AI name AI visibility agencies? https://trygoppa.com/study/agencies — CC BY 4.0.
No agency is named for a poor result. We publish the counts, the countries, the categories and the method; the identity of individual agencies stays out of it. The ones we do name are the ones that appear.