We asked four AI systems who the best agencies are. Here is what decided the answer
Asking the same category questions across ChatGPT, Perplexity, Gemini and Google AI Overviews returns overlapping but not identical shortlists. The companies named consistently were rarely the ones ranking first in ordinary search. What they shared was presence in third-party listings, specific published facts such as prices and timelines, and a clear description of what they do repeated identically across the web.
We ran a small experiment: the same questions a buyer would ask, put to four systems, with the named companies recorded.
This is not a study. It is an afternoon of testing, reported honestly, because it is more useful than another article explaining what generative engine optimization means in theory.
The method
Twenty questions of the kind a buyer actually types. Not “best digital agency” but the real ones:
- Who can build an ecommerce site for a small furniture brand?
- Which agencies work white label for other agencies?
- I need product photography and a website — can one company do both?
Each question asked in ChatGPT, Perplexity, Gemini and Google AI Overviews. Company names recorded. No prompting tricks, no follow-ups.
What came back
The shortlists barely overlap. A company named consistently by one system was frequently absent from another. Anyone testing a single system and reporting a result is measuring that system, not visibility.
Search ranking predicted very little. Several companies dominating ordinary search results for their category were named rarely or not at all. Several named repeatedly were nowhere near the first page.
Directories carried surprising weight. Answers leaned on Clutch, DesignRush, G2 and similar aggregators noticeably more than on the companies’ own websites. A well-maintained profile on one of those appeared to matter more than a beautifully designed homepage.
Specific facts got quoted directly. Where a company published a price, a turnaround or a constraint, that figure appeared in the answer, sometimes verbatim. Companies whose sites spoke only in adjectives were described generically or not mentioned.
What the named companies had in common
Three things, consistently:
- Presence in third-party listings, maintained rather than abandoned after signup.
- Specific published facts — prices, timelines, capacity, geographic limits. Things that can be quoted without interpretation.
- One consistent description of what they do, repeated near-identically across their site, their profiles and their listings.
None of those requires a budget. All three require deciding what you actually are and saying it the same way everywhere, which turns out to be harder than it sounds.
What we changed on our own site afterwards
We published prices for white label work instead of “contact us for a quote”. We stated the delivery windows. We wrote plainly that we take two or three agency projects at a time, and that our live streaming is regional.
That last one is a limitation, published deliberately. A system trying to decide whether to recommend us for an event in Chicago should have enough information to decide not to.
Try it yourself
It takes an afternoon and costs nothing. Write your twenty questions, run them across four systems, record who gets named.
If you are not in the answers, you now know. If you are, you know which questions and which systems — which is the only baseline worth measuring against next month.
Common questions
Do different AI systems give different answers?
Yes, and the overlap is smaller than you would expect. A company named consistently by one system is often absent from another. Testing a single system and drawing conclusions from it is the most common mistake in this area.
Does the answer change over time?
Yes. Ask the same question a month apart and the shortlist shifts. This is why a one-off audit has limited value and why the same fixed question set has to be re-run on a schedule to mean anything.
Can a small company get named?
It happens more than you would think, because these systems are not weighting purely by size or backlink count. A small company with specific published facts and consistent third-party presence can appear alongside far larger ones.
