GEO agencies in AI answers: who do the machines actually name?

We queried four answer engines: ChatGPT, Gemini, Claude and Google AI Overviews. In five languages — Hungarian, English, German, French and Polish — we asked pre-registered, unbranded category questions, 55 observations in total. The questions were the kind a real client asks: who can help my company appear in AI answers?

Every mention was recorded and published. The raw data — per question, per engine, with language and position — is available as open, downloadable CSV on the research page. Anyone can re-run and verify it.

One important exclusion: we conducted the measurement, so AVE Studio does not appear in this list — we do not rank ourselves, due to the conflict of interest. Our own result is public on the research page.

The list shows who the engines mention. Not who does good work.

The two are not the same. Being named feeds largely on what the engine finds on the open web: service pages, definitional articles, lists. Whoever publishes a lot, in well-structured form, about their own field gets named more often — regardless of the quality of their client work. The list therefore measures visibility, not quality. Read it that way.

And one more limit: generated answers vary. This is the result of one run, on one given day. The numbers show magnitude and pattern, not eternal truth.

For the Hungarian-language and Central European questions, the engines named the following agencies (by mention count):

• Peak Ace — 14 mentions, from all four engines, in three languages; the Berlin agency is a recurring name in the regional questions too
• TUYA Digital — 5 mentions, from two engines
• Delante — 5 mentions, from two engines, in three languages
• Marketing21 — 4 mentions, from two engines
• 'Vantgard Digital — 3 mentions, from three engines, for Hungarian questions
• HD marketing — 2 mentions, from two engines, for Hungarian questions
• AppLabx — 2 mentions, from one engine
• CRS AI Marketing & SEO Ügynökség, BP Digital, Contented, Tartalom Design, Optimist — 1 mention each

For the English and local-language European questions, partly different names come up:

• Eskimoz — 9 mentions, from all four engines, for English and French questions
• Promoovoir — 4 mentions, from three engines, for French questions
• Blue Array, Animalz, Siege Media, Onely, Keyweo — 2 mentions each
• Rise at Seven, WebFX, Directive Consulting, Single Grain — 1 mention each

Three things stand out.

First: the four engines give four different lists. There is hardly a name that every engine mentions — there is no single answer to "who is visible in AI"; it has to be measured engine by engine.

Second: being named is strongly language-dependent. Whoever is visible in the Hungarian questions is typically absent in the German ones — and the other way around. Multilingual presence is separate work; it does not come for free.

Third: the engines largely work from what the players publish about themselves — service pages, definitional articles, case studies. The list therefore reflects, to a large degree, who has written up their own category well for the machines.

One measurement run, on one fixed day, with 55 unbranded observations. The sample is small, the answers are variable, and the question set — however carefully built — does not cover the full space of buyer questions. We will re-measure at the next major model generation and publish the update here. If you are an agency on the list and want a correction, get in touch — the measured data is public, and we are happy to show the reasoning.

If you are a client: use the list as a starting point, not a verdict. When choosing, demand the things worth demanding of any provider — we wrote a separate article about that.

If you are an agency: your own number is measurable. With the same method, we measure when and where the machines say your name — and if they don't, why not.