ChatGPT Names a Clear Brand Leader in 15% of Categories. The Rest Are Up for Grabs.

MarTech reported a Semrush study on July 23 that tracked how ChatGPT answers 1,094 US product categories, five buyer questions each. In only 15.2% of them did a single brand clearly own the answer. In 53.7%, no brand led at all.

Across the categories that had any movement, leadership changed nearly 2,000 times over the study.

For a brand that just told its board it “shows up in ChatGPT,” that number is the problem. Showing up in one answer and owning the category are different things, and most brands have the first without the second.

The gap widens exactly where it hurts. The most-searched topics accounted for roughly 98% of AI search volume, and only 11.3% of those high-demand categories had a clear owner, against 19% of the quieter ones. The categories worth winning are the ones least settled.

None of this tracks with the scoreboard brands are used to. Kevin Indig, the growth advisor who partnered on the study, said the old metrics don’t explain the outcome:

“Traditional SEO metrics aren’t enough to explain who owns a topic. While they play their role, there’s more to it.”

The clearest sign of the decoupling: only 21% of the most-cited domains in a category were also the most-mentioned brand in the answers. The site ChatGPT links to and the brand it names are frequently not the same, which means a page-one ranking and a citation are no longer the same asset. You can be the source the model reads and still not be the name it says.

Raouf Douihech, an independent SEO and generative-search consultant who publishes a working guide to the discipline, opens it with a warning most agency decks leave out:

“this field is not stable. The platforms change their mechanics constantly, and studies land every week that refine or contradict the previous ones.”

The mechanism behind the churn is simple once you stop picturing a ranked list. A search engine returns ten fixed blue links; the order changes slowly and you can watch your position. A model generates a fresh answer each time, weighing what it has read, what it was asked, and what it happens to surface in that run. When your lead over the runner-up is a point or two, the next regeneration can hand the mention to someone else. Ownership is real only when the lead is wide, and in most categories it isn’t.

That reframes the AEO spend a lot of brands greenlit this year. Optimizing to “get cited” treats AI visibility like a ranking you climb once and hold. The study describes a surface that reshuffles constantly and where citations and mentions come apart, which is the same disconnect we found when AI visibility stopped tracking referral traffic and when search teams started chasing an AI-search score their vendors couldn’t define. A metric that moves 2,000 times isn’t a target you set a quarterly goal against.

So treat AI visibility as a position to defend, not a milestone to hit. Three things follow for a brand budget. Measure share of mention over weeks, not a single flattering screenshot, because one answer proves nothing about the next. Separate the two goals your agency is probably billing as one: being the cited source is a content-and-links job, being the named brand is a reputation-and-demand job, and the study says they don’t move together. And spend against the high-demand categories precisely because they’re unsettled; a wide-open topic is cheaper to win now than after someone else’s lead gets wide. The brands that treat the answer box as weather, not a scoreboard, are the ones that will still be named in it next quarter.