Are AI answers about a market dominated by a few big brands?

In our scan of 74 buyer questions, no. 273 different companies were named across the answers, and the single most-mentioned one appeared in only 20% of them. There was a modest cluster of frequently-named companies but no dominant winner — the field was fragmented, not concentrated in a few hands.

Why does fragmentation in AI answers matter for my company?

It makes presence contestable: there is no incumbent owning the answers to unseat, since even the most-visible company missed four of every five. But a single mention is nearly worthless — 79% of named companies appeared only once. What separates real presence from a coincidence is showing up consistently across the questions that matter.

What should I measure given how fragmented AI answers are?

Coverage, not rank. Because no company owns the answers, beating an incumbent is the wrong goal. The right question is: across the specific buyer questions that matter to you, in how many do you appear, and how consistently? Repeated appearance across a fixed question set is the honest metric — not the existence of one citation.

AI answers don't crown a winner: what 273 companies across 74 questions reveal

Guide · AI Visibility · 4 min read · last verified 2026-07-23

Reviewed before publication Editorial board Independent commercial review

In shortWe expected a few incumbents to dominate AI answers about their market. Our own scan of 74 buyer questions found the opposite: 273 different companies named, 79% of them just once, and the most-visible company in only one answer of five.…

When we measured how companies actually appear in AI-generated answers about their own market, we expected to find a handful of incumbents dominating the results. The data said the opposite. Across 74 real buyer questions in our own research, 273 different companies were named in the answers — and 79% of them appeared exactly once. Even the single most-mentioned company surfaced in only one answer out of five. AI answers about a market are not winner-take-all. They are fragmented, and that changes what it takes to compete in them.

What did the numbers actually show?

The scans covered 74 buyer questions across a set of B2B markets, and every company that any answer named was recorded from the source pages behind it. The result was a very long list. According to our own scan data, 273 distinct companies were mentioned across those 74 answers — more than three companies named per question, with almost no repetition. In our scan, the most-visible company in the entire set still appeared in just 20% of the answers; the top ten companies together accounted for only about a fifth of all mentions. Below that thin head, the field collapsed into a tail: 215 of the 273 companies — 79% — were named in a single answer and never again.

This is not the shape people assume. The intuition, borrowed from search, is that a few strong brands own the results and everyone else fights for scraps. In these scans, there were no owners. There was a modest cluster of frequently-named companies, and beneath it an enormous, flat tail of near-invisibility.

Why are AI answers so fragmented?

A reasonable hypothesis — and we label it a hypothesis, because it describes model behavior we cannot observe directly — is that answering systems optimize for relevance to the specific question, not for a stable ranking of the market. A search results page tends to return the same strong domains across related queries. An answer composed for one narrow question pulls from whatever pages best fit that question, which may be a different set entirely from the pages that fit the next one. The consequence is breadth: many companies each catch a specific question, few companies catch many.

Whatever the mechanism, the measured effect is what matters. Presence in AI answers is distributed thinly and widely, not concentrated in a few hands.

What does fragmentation mean if you are trying to be found?

It cuts two ways, and both are more encouraging than the winner-take-all story.

First, presence is contestable. You are not trying to unseat a dominant incumbent who owns the answer, because no such incumbent exists — the most-visible company in our set still missed four of every five answers. The competition is not a wall; it is a crowd. Earning your way into the answers for the specific questions you can credibly answer is a realistic goal, not a moonshot.

Second, and less comfortably, a single mention is worth very little. Being one of 273 companies named once is not presence; it is a coincidence. The 79% long tail is full of companies that got a token appearance and no consistent footing. What separates the thin head from that tail is not a single lucky citation — it is showing up repeatedly, across the cluster of questions that matter in your category. Consistency, not a one-time hit, is the thing that is actually scarce.

How should this change what you measure?

If the field were concentrated, the sensible metric would be rank: are we beating the incumbent? Because the field is fragmented, rank is the wrong question. The right one is coverage: across the specific set of buyer questions that matter to us, in how many do we appear at all, and how consistently? A company can be "mentioned in AI" — technically true, once, somewhere — and still have no real presence. The honest metric counts repeated appearance across a fixed set of questions, not the existence of a single citation.

That reframes the goal from a race you were told you had already lost into a build you can actually start. In a market where 273 companies split the answers and almost none of them show up twice, the durable advantage goes to the company that becomes one of the few that consistently does.

How far should a five-scan sample carry?

This is what five delivered scans across 74 questions found, not a law of every market. Five companies is a small sample, and the specific markets shape the specific numbers. What we would not expect to change with a larger sample is the direction of the finding: that the number of distinct companies named is high, the repetition is low, and no single company owns the answers. If anything, a broader sample of markets would likely widen the tail, not narrow it. The exact figures belong to these scans; the shape is the part worth carrying forward.