“Our category’s already locked up by the big players, isn’t it.”
Say that in a planning session and the room tends to go quiet, because everyone half believes it.
And the person who said it isn’t being lazy. Ask ChatGPT a category question and the big names really do line up in the answer.
I’ve watched ten answers in a row come back without our name in them and quietly closed the tab, so I’m not throwing stones here.
But “locked up” and “nobody’s claimed it yet” look identical from that tab. Have we ever counted the difference? Somebody has — 1,094 categories’ worth.
Has The AI Already Picked A Winner In Your Category?
Mostly it hasn’t. Across 1,094 US categories checked in ChatGPT, 15.2% had a clear owner and 53.7% had no one you could call an owner at all.
Another 31.2% had what the study calls an emerging leader, which is a race in progress rather than a result. So the categories where the thing is genuinely settled are that 15.2% — about one in six or seven.
A category owner here is a brand that comes back reliably when you put that field’s questions to the model. The work is Semrush’s, published 20 July 2026 (the study).
- More than 50,000 brands, and 1,094 categories in the US
- January to June 2026, ChatGPT and nothing else
- Five representative prompts per category: definition, comparison, alternatives, use case, and purchase consideration
Semrush sells tracking tools for exactly this, so “you should measure at the category level” is a conclusion that pays them. That’s not a reason to throw the numbers out; it is a reason to hold them at arm’s length while you read.
Even The Winners Only Lead By 1.3 Points
Now for how the rankings actually moved over those six months.
Where the top spot changed hands, the median lead was 1.3 points. Where it held, 2.9 points. And clear owners kept their position in 90.4% of month-over-month comparisons.
Look at 90.4% alone and the incumbents look immovable. The thing worth sitting with is what that immovability rests on: a handful of points.
Think of a shelf in a convenience store, one product nudged a single row forward. Restock it tomorrow and a different name is at the front.
If you know the gap in your own category, you know whether it’s worth funding. The 2.9-point median gives you something to hold that number up against.
Does Branded Search Tell You Who Owns The Category?
Only about half the time, which is less useful than it sounds. Branded search volume predicted the category owner in 55.7% of cases; Semrush’s own Authority Score managed 52.5%, and organic traffic 48.4%.
Branded search is the count of people typing your brand name into a search box, and it’s a decent read on how far your name has spread. As a guide to who the model treats as the authority, it’s barely better than a coin.
There’s a second pairing worth separating too. Mentions (how often a brand’s name comes up) and citations (how often a domain gets used as a source) correlate at -0.229 — very slightly negative.
And the most-cited domain in a category was also the most-mentioned brand only 21% of the time.
Since mentions and citations move independently, they belong in two columns rather than one total. I’ve written before about folding AI visibility into the same team as search (here); merging the team still leaves you adding columns to the report.
How Far This 53.7% Travels
ChatGPT only. Gemini, Perplexity and AI Overviews are all outside the data.
The 1,094 categories are US ones, so nothing here is a measurement of Japanese-language or Japanese-market behaviour.
Five prompts per category is a clean design, though it isn’t weighted to match how demand is actually distributed across the questions people ask.
“Owner” and “emerging leader” are Semrush’s own thresholds. Put their percentage next to another vendor’s and you’re comparing two different definitions, not two readings of the same thing.
And the branded-search relationship is a correlation. Nothing in the study says pushing branded search up will make a model treat you as the authority.
Where I’d land: 53.7% is a good reason to go count your own category, and a bad substitute for having counted it. It describes 1,094 American categories, not yours.
Pick One Category And Count The Gap To First Place
The useful part of this study isn’t any single percentage. It’s the unit of measurement.
Ask how visible the whole company is in AI answers and the places you’re winning cancel out the places you’re losing, so the number comes back grey and unusable.
The concrete version is small: choose the one category you actually want to win, and measure how many points separate you from whoever leads it today. A few points and it’s a candidate for investment. Twenty and you probably want a different category first.
Then hold it steady — same questions, same category, once a month. Industry reports won’t do this part for you, because they’re not measuring the category you sell into.
Count it a few times and the screen that looked full of big names starts showing gaps. Go get one category’s number and take it back into the meeting.
Sources
- Margarita Loktionova (Semrush), “AI visibility is a topic-level game: A study of 50,000 brands in ChatGPT”, published 20 July 2026, semrush.com (over 50,000 brands across 1,094 US categories, January-June 2026, ChatGPT only, five representative prompts per category: definition, comparison, alternatives, use case, purchase consideration. A clear category owner existed in 15.2% of categories; 31.2% had an emerging leader; 53.7% had none. Median lead of 1.3 points where the top spot changed hands, 2.9 points where it held; clear owners retained position in 90.4% of month-over-month comparisons. Branded search volume predicted the owner 55.7% of the time, Authority Score 52.5%, organic traffic 48.4%. Mentions and citations correlated at -0.229, and the most-cited domain was the most-mentioned brand in only 21% of categories. Semrush sells AI-visibility tracking tools; “owner” and “emerging leader” are its own definitions, and the branded-search relationship is correlational only.)
