Short answer. Across 1,094 tracked US categories in ChatGPT, only 15.2% had a clear brand owner, 31.2% had an emerging leader and 53.7% were unsettled. Where an owner does exist, it holds position in about nine months out of ten. So the strategic question is not "how do we win AI search" but "which of those two categories are we in" — because the answers are opposites, and the most encouraging detail is counter-intuitive: the highest-volume topics are the ones least likely to have an owner.
Category ownership in AI answers behaves nothing like a search ranking. Nothing in an engine records that brand X leads category Y; ownership is an emergent property of retrieval repeating across the many separate questions that make up a topic. That changes both how it is measured and how it is won.
This piece works from the largest public study of the question, then sets out how to find out which game your category is in — an answer available before a single page is written.
How many categories actually have an owner?
Semrush, with Kevin Indig for Growth Memo, mapped brand presence across 1,094 US categories in ChatGPT between January and June 2026, covering more than 50,000 brands, 220,000 domains and 600,000 citations.
| Category structure | Share of 1,094 categories |
|---|---|
| Clear owner | 15.2% |
| Emerging leader | 31.2% |
| Unsettled | 53.7% |
Clear owners kept the top spot in 90.4% of month-over-month comparisons. But the finding worth reading twice is the popularity split: only 11.3% of the most popular half of topics had a clear owner, against 19% in the less popular half — and that popular half carried 98% of AI search volume.
That inverts the usual assumption. The most contested, highest-volume topics — the ones buyers actually ask about — are in aggregate the least settled.
How different are the two situations?
| Unsettled category (53.7%) | Owned category (15.2%) | |
|---|---|---|
| What the engine does | Names a rotating cast; no consistent leader | Names the same brand across months |
| Realistic goal | Become the emerging leader | Be a credible second source |
| What decides it | Coverage breadth and consistency over months | Displacing an incumbent with a 90.4% hold rate |
| Time horizon | Quarters | Years, or a market event |
| Leading indicator | Rising share of runs across a fixed question set | Appearing as an alternative in comparison answers |
| Wrong move | Waiting for certainty before publishing | Buying a programme that promises displacement |
The margins matter as much as the labels. In the same study, where leadership changed hands the median lead was 1.3 percentage points; where it held, 2.9. A category leader with a thin margin is not secure, and a challenger within a couple of points is genuinely in contention — a narrower gap than most brands assume when they look at an incumbent's brand recognition.
Why is ownership topic-level rather than brand-level?
Because it is won the same way it is measured: question by question.
- Each question retrieves independently. A brand cited on one question in a topic is not thereby favoured on the next one. There is no accumulated authority score being carried forward inside the answer.
- Consistency across a topic's questions is what looks like ownership. Being named on four of a topic's twenty questions is not ownership. Being named on fourteen is, even if never on the same day twice.
- Adjacent questions compound. Coverage that answers the definition, the comparison, the cost, the failure modes and the alternatives enters more retrieval pools than coverage that answers only the flattering question.
- Third-party corroboration does most of the work. Roughly 85% of AI references point away from brand-owned domains, and the top 15 domains account for roughly 68% of all citations produced by the five major engines.
That last point is the uncomfortable half. The domains doing most of the citing are a small set of large platforms, and no amount of publishing on your own site changes what those platforms say about you. Category ownership is contested substantially on ground you do not control.
Where does a challenger have the best chance?
- Engines with more citation slots. ChatGPT cites about 15 sources per answer against Gemini's 3, on the Semrush AI Visibility Index built from 126 million US prompts recorded January to April 2026. A five-fold difference in slots is a five-fold difference in room for a non-incumbent.
- Unsettled categories, which is most of them. The 53.7% figure is the single most encouraging number in this data.
- High-volume topics, counter-intuitively. They are less likely to have a clear owner than the quiet tail.
- Questions the incumbent answers badly. Comparison, limits, cost and failure-mode questions are chronically under-served, because incumbents avoid writing them.
- Non-English markets. Coverage in most categories is thinner in every language other than English, and an incumbent's advantage is usually an English-language advantage rather than a category one.
How do you find out which game you are in?
- Define the category as a question set, not a keyword. Fifteen to forty questions a buyer would actually ask an assistant, covering definition, selection, comparison, cost, risk and alternatives.
- Run them repeatedly and record every brand named, not only yours. The distribution of competitor names is the ownership signal; your own rate is only one column of it.
- Compute concentration, not rank. If one brand appears in more than about half of runs across the set, you are in an owned category. If the top brand sits under a third and the field is wide, you are in an unsettled one.
- Check the margin. A leader two or three points clear of the next brand is beatable on this evidence. A leader far clear is not, on any reasonable budget.
- Re-run monthly and watch the hold rate. Month-over-month persistence is what tells you whether the leader is established or is simply this month's draw.
This costs nothing but time and repetition, and it should be answered first, because it determines whether the sensible plan is a two-quarter push or a three-year presence programme. Those are different budgets and different teams.
What this data does not say
- It is ChatGPT, and it is the United States. Category structure elsewhere is not measured here, and engines differ enough that it should not be assumed to transfer.
- Six months is a short series. A 90.4% month-over-month hold rate across a six-month window is strong evidence of stickiness, not proof of permanence.
- Ownership is not revenue. Being the named brand in an answer is influence over a reading buyer, not an attributed sale.
- The correlation with traditional SEO metrics is weak. In the same study, category owners had higher organic traffic in only 48.4% of pairwise comparisons and a higher authority score in 52.5% — both close to a coin flip. Owning a category in AI answers is not a by-product of ranking well.
How Lifewood approaches this
Lifewood answers the category-structure question before scoping content, because the answer changes what should be bought. A fixed question set is run repeatedly with every brand named recorded, not just the client's, and the output is a concentration figure and a margin rather than a rank.
Where a category comes back unsettled — most do — the plan is breadth across the topic's questions, including the comparison, cost and failure-mode questions incumbents avoid. Where a category comes back owned with a wide margin, the honest recommendation is second-source presence and a longer horizon, and Lifewood says so rather than selling a displacement programme it cannot deliver.
Because the incumbent's advantage in most categories is an English-language advantage, non-English markets are usually where the concentration figures are softest. Running the same question set natively per market rather than translated is what makes that comparison meaningful, and 50+ languages across 40+ delivery centres in 30+ countries is what makes native authorship practical. See how to measure AI visibility without fooling yourself and where AI answer engine citations go.
Sources and further reading
- Semrush with Kevin Indig, AI visibility is a topic-level game: a study of 50,000 brands in ChatGPT, January–June 2026 — the source of every ownership, popularity-split, margin and correlation figure above.
- Semrush, 2026 AI Visibility Index, 126 million US AI search prompts — citations per answer by engine.
- Omnibound, Answer Engine Optimization statistics 2026 — third-party citation share.
- AI Platform Citation Source Index 2026, synthesis of six studies covering 680 million citations — domain concentration.