How AI Picks an Agency: 608 Answers, 66 Brands, Measured

Being named is far more concentrated than anyone expects

66 different agencies were named at least once across 608 answers, which sounds like a crowded field. It is not. Those 66 brands share 1,302 namings between them, and the top five take 40% of them. Only 5 brands are named in more than one answer in ten. 26 of them, well over a third of everyone who appears at all, turn up in fewer than one answer in a hundred.

A concentrated category is a winnable one. The gap between the median brand and the top five is not reputation or age. It is how often each one appears in the sources these models read before they answer.

What the models actually read

Every answer was traced back to the sources cited to produce it. reddit.com came first by a distance, cited in 225 answers, roughly 52% more often than the next domain and more often than any vendor website in the category, including the sites of the agencies that get named most.

This is the practical finding. Most effort in this category goes into a company's own website, and a company's own website is not what decides the answer. Getting named means being present, accurately, in the sources above.

What this changes about getting named

Three things follow from the data, and they are the same three things we do for clients.

• Measure the question, not the page.

  Ranking for a keyword and being named in an answer are different outcomes with different causes. Until you track which specific buying questions name you, on which engine, you are guessing.

• Work the sources, not just the site.

  The most-cited domain in this category is a community site, not a vendor. Presence in the sources an engine reads is what moves the answer, and that work happens off your own website.

• Expect it to be winnable.

  With 5 brands holding more than a tenth of the answers each and the median brand at 1.48%, the distance between invisible and named is smaller than it looks.