Who Owns Your Category in AI Search? What 1,094 Categories Show
Who Owns Your Category in AI Search? What 1,094 Categories Show
Quick answer: In a study of 1,094 US categories tracked in ChatGPT from January to June 2026, only 15.2% had a clear brand owner — a brand named in at least four of five related buyer prompts with a five-percentage-point lead over the runner-up. Another 31.2% had an emerging leader, and 53.7% were unsettled, meaning no brand appeared in even three of the five prompts [1]. That does not mean those categories are empty. A separate academic preprint measuring 3,750 responses across three models found true competitive vacuums in only 8.0% of queries, with moderate concentration overall [2]. Put together: AI almost always names someone, but very few brands get named consistently across a whole topic. Category ownership is a coverage problem, not a ranking problem — and coverage is still available in most categories.
If your AI visibility report shows you appearing in a handful of prompts, you do not yet know whether you own anything.
---
Why does winning one prompt tell you so little?
Because buyers do not ask one question.
Someone evaluating payroll software asks what it is, how two options compare, what the alternatives are, whether it fits their situation, and what to buy. That is five different prompts inside one buying decision, and a brand can win one and vanish from the next four.
This is the structural reason single-prompt tracking misleads. The Semrush study built each category as a cluster of five representative prompts covering definition, comparison, alternatives, use case, and buying question — then asked whether any brand held presence across the cluster [1]. Measured that way, presence turns out to be far patchier than a single-prompt check suggests.
The practical consequence: a dashboard showing "you appear in 40% of tracked prompts" can describe a brand that owns nothing, because the 40% may be scattered across unrelated topics rather than concentrated inside one.
---
How many categories actually have an owner?
Few. Here is the full breakdown across the 1,094 categories analyzed [1]:
| Category status | Share | What it means |
|---|---|---|
| Clear owner | 15.2% | One brand appears in at least 4 of 5 prompts with a 5-point lead over the runner-up |
| Emerging leader | 31.2% | One brand appears in at least 3 prompts but without the required lead |
| Unsettled | 53.7% | No brand appears in even 3 of the 5 prompts |
The counterintuitive part is where the ownership sits. Splitting the categories by AI search demand, only 11.3% of the high-demand half had a clear owner, against 19% of the low-demand half. And the high-demand half accounted for 98% of the AI search volume in the sample [1].
So the pattern runs opposite to traditional search, where high-volume terms are usually the most locked down. Combining the two figures, roughly 87% of the AI search volume in this sample sat in categories with no clear owner.
One scope note before anyone reallocates a budget on this: the study draws on Semrush's own AI Visibility Toolkit data, produced in partnership with Kevin Indig and Growth Memo. It is vendor research with a disclosed method, not an independent audit, and it covers US categories in ChatGPT only [1].
---
Does "no owner" mean the category is empty?
No, and this is the distinction that decides whether the finding is useful or misleading.
A separate study approached the same question from the model side. It ran 250 brand-free category queries across 50 brands and five industries on three models — GPT-5.2, Gemini 3 Flash, and Perplexity sonar-pro — repeating each query five times for 3,750 total responses. It then measured concentration directly [2].
Two results matter here. Mean concentration came in at a Gini coefficient of 0.28 (95% CI 0.16 to 0.41), below the 0.60 power-law threshold the author set — moderate, not winner-takes-all. And competitive vacuums, defined as categories with no single leader, appeared in only 8.0% of queries. In most cases the models named at least one of the sampled brands [2].
Set the two studies side by side and the apparent contradiction resolves:
| Question | Finding | Source |
|---|---|---|
| Does AI name brands in a given answer? | Yes, almost always — vacuums in 8.0% of queries | Preprint, 3 models [2] |
| Does one brand hold presence across a whole topic? | Rarely — 15.2% of categories | Vendor study, ChatGPT [1] |
The gap between those two facts is the opportunity, and it is a narrower opportunity than "85% of categories are wide open" suggests. Competitors are already being named in individual answers. What almost nobody has is consistency across the cluster.
This preprint carries its own limits. It was submitted in June 2026 and is under review at a journal, so it has not completed peer review. It sampled 50 brands across five industries, which is small, and the author frames the three metrics as exploratory and the results as sitting in tension with a strong winner-takes-all narrative "within the scope studied" [2]. Treat it as a useful counterweight with a small sample behind it.
---
What does it take to count as the owner?
The Semrush thresholds give you something a visibility percentage does not: a definition you can aim at [1].
- Appear in at least four of five prompts in the category
- Hold the highest share of mentions
- Lead the runner-up by at least five percentage points
Two things to keep straight about these numbers. They are that study's operational definitions, chosen to separate signal from noise in its dataset — not thresholds any AI platform publishes or enforces. And ownership there was measured by brand mentions in the answer text, not by source citations, on the reasoning that readers act on what the answer says [1].
As a working target, the definition is more actionable than most GEO goals because it names the unit of work. The thing you are trying to cover is a question cluster.
---
Why doesn't strong SEO decide this?
Sitewide strength does not settle it. Ranking across the specific question cluster looks far more relevant.
The study compared each category owner against the runner-up on three domain-level metrics. Owners had higher branded search volume in 55.7% of pairs, higher organic traffic in 48.4%, and a higher Authority Score in 52.5%. Only branded search volume was statistically significant, and the edge was modest [1].
Organic traffic below 50% is the number worth sitting with. In that dataset, the brand with more organic traffic lost the category slightly more often than it won.
Kevin Indig, who co-produced the research, draws the line carefully:
"Treat that as a hypothesis, not a finding. What the data actually shows is: traditional SEO metrics aren't enough to explain who owns a topic. While they play their role, there's more to it."
An independent meta-analysis helps locate where that line falls. Cyrus Shepard scored 23 factors associated with AI citations by drawing on 54 experiments, patents, and case studies, weighting each by repeatability, strength of evidence, and official platform support. Domain authority scored 5.0, with studies finding a relationship but typically a weak one. Search rank scored 9.4, fan-out rank 9.3, and topic cluster ranking 8.9, on the finding that a site ranking across multiple related queries gains a compounding probability advantage [4].
The two sources arrive at the same place from different directions. Sitewide authority is weakly associated with citation and does not explain category ownership. Ranking across the cluster of related questions is where the association concentrates — which is the same behavior the four-of-five coverage threshold describes.
Shepard states the limit of his own framework plainly: the 23 items are "features correlated with AI citations across multiple studies," not ranking factors in the traditional sense [4]. The direction is well replicated. The mechanism is not established, and no source reviewed here isolates it causally.
---
Once you own a category, do you keep it?
Usually, if the lead is wide enough.
Clear category owners held first place in 90.4% of month-over-month comparisons. Among emerging leaders and unsettled categories, the top brand changed in 1,950 of 5,470 comparisons [1].
The margin is what separates the two outcomes:
| Outcome | Median lead |
|---|---|
| Leader flipped the next month | 1.3 percentage points |
| Leader held | 2.9 percentage points |
That gives you a defensive read as well as an offensive one. A one-point lead behaves like a coin flip; a three-point lead behaves like a position. And if the current leader in a category you care about is sitting on a narrow margin, that is the cheapest moment to contest it.
The study is explicit that it cannot say why positions change — shifts could reflect changes in ChatGPT's sources, changes in answer generation, or real changes in brand relevance. What the data supports is a pattern about margin, not a mechanism [1].
Six months of monthly data is also a short window for a retention claim. Read 90.4% as an observation from that window, not a durability guarantee.
---
Does owning it in ChatGPT mean owning it in Gemini?
No. This is the most consistently replicated finding across the sources here, and the one most likely to break a measurement program.
The preprint found cross-model agreement on the top-recommended brand of just 41.6%. A top position on one model did not reliably hold on another [2].
An industry index published in August 2026 reached a compatible result in software specifically. GetIntel ran roughly 80 buyer-phrased questions in each of 126 software categories covering 1,825 brands, putting them to the consumer ChatGPT and Gemini apps rather than the developer APIs, and analyzed 9,978 usable answers collected on August 6, 2026. The two assistants named a different #1 product in one of every three categories [5].
The ghost citations study adds the mechanism. Across 454 prompt-and-domain combinations tested on multiple engines, the engines disagreed on whether to name the brand in 100 cases, or 22% [3]. Indig's summary of that dataset:
"There's almost no overlap between which brands ChatGPT cites and which ones Gemini names for the same prompt. These are different behavioral systems. Treat them that way."
A separate Semrush analysis of more than 1,200 brands across 22 verticals found only 36 appearing in the top 100 most-mentioned list on every platform in every month of its window [6].
The operational conclusion is narrow and firm: a blended AI visibility score averages away the divergence that matters. Category ownership has to be measured per engine or it is not measured at all.
---
Are you measuring mentions or citations?
These are different outcomes, and for category ownership they point in nearly opposite directions.
In the ghost citations dataset — 3,981 domain appearances across 115 prompts, 14 countries, and four engines — 61.7% of appearances were citations without a brand mention. Only 13.2% were both cited and mentioned, and 25.1% were mentions with no citation. Overall, 74.9% of appearances included a citation while just 38.3% included a mention [3].
Engine behavior diverges sharply on this axis:
| Engine | Mention rate | Citation rate |
|---|---|---|
| Gemini | 83.7% | 21.4% |
| ChatGPT | 20.7% | 87% |
Source: 3,981 domain appearances, four engines [3]
The topic-ownership study found the same split inside its own data: only 21% of the most-cited domains in a category were also the most-mentioned brand, and the two correlated slightly negatively at −0.229 [1].
Worth noting that two research teams working independently landed on the same measurement choice. The preprint's Category Ownership Index is defined as a brand's share of mentions within a category [2], and the topic-ownership study also measured ownership by mentions in the answer text [1]. Neither used citations as the ownership unit.
So a citation-only report can show a rising line while your brand goes unnamed in the answers buyers actually read. Both signals are worth tracking, but they answer different questions, and treating one as a proxy for the other will misprice the work.
---
How do you go after a category?
The sequence below follows what the evidence supports, in the order the decisions arrive.
- Pick two or three categories. Choose subject areas where you already have some presence, existing content, and a real product angle. Spreading across every topic dilutes coverage, which is the exact thing ownership requires.
- Write out the five-question cluster for each. Definition, comparison, alternatives, use case, buying question. This is the unit of work, and it is where most content programs turn out to have gaps — usually on comparison and alternatives. Match the format to the question type while you are at it: intent-format match scored 9.0 in Shepard's framework, with comparison-style queries favoring tables and lists and how-to queries favoring numbered steps [4].
- Check where you sit per engine, per category. Not a blended score. A category you lead in ChatGPT may have a different leader in Gemini one time in three [5].
- Track mentions and citations separately. If you are cited without being named, the fix is usually clearer brand references and comparison sections that force the model to name the players, rather than more content [3].
- Work the off-site layer. Models name brands they encounter consistently across the web. Unlinked brand mentions in community discussions, review sites, industry publications, and expert commentary all feed that familiarity [1].
- Watch the margin, not the direction. A lead under two points is contestable in either direction. Push until the gap is wide enough to hold.
Step 3 is where most teams stall, because the measurement has to be shaped like the problem. Three capabilities decide whether you can see category ownership at all:
| Capability | Why ownership needs it |
|---|---|
| Per-engine breakdown | Cross-model agreement on the top brand was 41.6%; a blended score hides which engine you are losing [2] |
| Category-level grouping | Ownership is defined across a prompt cluster; prompt-by-prompt lists cannot show coverage |
| Mentions tracked apart from citations | 61.7% of appearances were citations with no mention [3] |
Innflows is built around this layer. It tracks brand citations and mentions engine by engine — ChatGPT, Google AI features, Perplexity, and Copilot alongside Chinese-language engines including DeepSeek, Doubao, Qwen, and Kimi — and audits whether AI crawlers can reach and parse the pages meant to support each category, including content structure and entity consistency. Once per-engine baselines exist by category, "do we own this topic" becomes a number you can review quarterly.
The audit half matters as much as the tracking half here. Building coverage across a five-question cluster does nothing if the pages carrying it are blocked at the CDN, dependent on JavaScript for their key facts, or inconsistent in how they name your brand and products.
---
Where this evidence stops
Four boundaries worth stating before this becomes a plan.
The ownership thresholds are one study's definitions. Four of five prompts and a five-point lead are analytical choices that produced a readable signal in one dataset. No platform publishes or enforces them.
Two of the four main sources are vendor research. The topic-ownership and ghost citations studies both use Semrush's own toolkit data, and the software index comes from an AI visibility vendor. Methods are disclosed in each case, which is why they are usable, but none is an independent audit.
The academic source is a preprint. It is under review, sampled 50 brands, and describes its own metrics as exploratory [2].
No source here establishes causation. These studies describe what ownership looks like and what fails to predict it. None isolates what produces it. Broader reviews of the field make the same point about GEO generally: it behaves as a probabilistic pipeline rather than a single ranking task with controllable inputs [7]. Anyone promising you a category position on a fixed timeline is selling something the evidence does not support.
---
FAQ
What counts as owning a category in AI search?
In the study that defined it operationally, a category owner is the brand with the highest share of mentions that appears in at least four of five related prompts and leads the runner-up by at least five percentage points. Only 15.2% of 1,094 categories met that bar; 31.2% had an emerging leader and 53.7% were unsettled [1]. These are that study's thresholds, not published platform rules.
If 85% of categories have no clear owner, are they empty?
No. Those categories mostly have brands appearing inconsistently rather than no brands at all. A preprint measuring three models found true competitive vacuums — categories with no single leader — in only 8.0% of queries, with moderate overall concentration at a mean Gini of 0.28 [2]. The opening is in consistency across a topic.
Does strong SEO mean I will own my category in AI search?
Sitewide strength is not enough on its own. Comparing owners against runners-up, owners had higher branded search volume in 55.7% of pairs, higher organic traffic in 48.4%, and a higher Authority Score in 52.5% — only branded search volume was statistically significant, and modestly so [1]. An independent scoring of 23 citation-associated factors puts domain authority at 5.0 while placing search rank at 9.4, fan-out rank at 9.3, and topic cluster ranking at 8.9 [4]. What appears to matter is ranking across the related questions in a category, not overall domain strength.
If I own a category in ChatGPT, do I own it in Gemini?
Assume not. Cross-model agreement on the top-recommended brand was 41.6% in a three-model study [2], and an index of 126 software categories found ChatGPT and Gemini naming a different top product in one of every three [5]. Measure each engine separately.
Should I track citations or brand mentions?
Both, separately. Across 3,981 domain appearances, 61.7% were citations with no brand mention, and mention rates ranged from 20.7% on ChatGPT to 83.7% on Gemini [3]. Within a category, only 21% of the most-cited domains were also the most-mentioned brand, and the two correlated slightly negatively [1]. A citation-only report can rise while buyers never see your name.
How long does it take to establish category ownership?
No source reviewed here measures a time-to-ownership figure, so any specific answer would be invented. What the data shows is that positions held with a median lead of 2.9 percentage points and flipped with a median lead of 1.3, and that clear owners retained first place in 90.4% of month-over-month comparisons across a six-month window [1]. That describes what a durable position looks like once you have it, not how long it takes to build.
---
Bottom line
Across 1,094 categories tracked in ChatGPT over six months, only 15.2% had a brand that showed up consistently across a topic's related buyer questions, and the highest-demand categories were less likely to have an owner than the low-demand ones [1]. That is a real opening, but not an empty field: models named at least one brand in the large majority of queries, and concentration measured as moderate rather than winner-takes-all [2].
The reframe worth taking away is the unit of work: ownership is measured across a question cluster, and it is held or lost by margin, with three points behaving like a position and one point behaving like a coin flip [1].
Two measurement corrections follow from the same evidence. Ownership has to be read per engine, because agreement on the top brand runs near 41.6% [2]. And mentions have to be read apart from citations, because 61.7% of appearances are citations where the brand is never named [3].
Pick two or three categories, map the five questions buyers actually ask inside each, and get an honest per-engine baseline before committing a quarter of content to any of them. If you want to see which categories currently cite you and which name you, engine by engine, start with an AI visibility audit and use the gaps to choose where to compete.
---
References
[6] - Semrush: 36 brands win AI visibility everywhere, 1,200 vanish on one — PPC Land