89% of AI Search Demand Has No Clear Brand Owner — What 1,094 Categories Reveal (2026)

Key takeaway: Tracking 1,094 US categories in ChatGPT answers for six months, only 15.2% had a clear brand owner, and 89.3% of estimated demand sat in categories with no clear owner (Growth Memo). That sounds like an opportunity — but the same data shows the other side: once a brand became the clear owner, it kept the top spot in 90.4% of monthly comparisons. The window is open, and it locks. This post covers why a citations-only strategy fails to produce brand mentions (the top-cited domain matched the top-named brand just 20.8% of the time), and what to measure now to claim a category (as of 2026-07-30).

---

Three-line summary

• Most AI answer categories have no clear owner yet. Of 1,094 categories, only 15.2% had a clear owner; 53.7% were open competition with no leader.
• But ownership hardens once established. Owners kept the #1 position at a 90.4% monthly rate — the opportunity and the lock-in live in the same dataset.
• Citations are the ticket, not the seat. The most-cited domain matched the most-named brand in only 20.8% of cases. Link- and citation-centric strategies alone do not earn the brand name in the answer.

What was measured — defining "category ownership"

The study tracked ChatGPT answers for 1,094 US categories from January through June 2026, using Semrush AI Visibility Toolkit data: 220K+ domains, 50K+ brands, and 600K+ citations (Growth Memo).

Element	Criteria
Prompts per category	5 types — definition · comparison · alternatives · use case · purchase decision
"Clear owner" definition	① highest share of brand mentions in the category ② named in at least 4 of the 5 prompt types ③ at least 5 percentage points ahead of the runner-up — all three at once
Platform	ChatGPT (US)

The notable part is that ownership requires consistency across prompt types, not a single winning question. A brand that only appears in comparison prompts, or only in definition prompts, is not an owner.

Results — 15.2% of categories owned, 89.3% of demand unclaimed

Segment	Share	Meaning
Clear owner exists	15.2%	One brand consistently dominates the answers
Open competition	53.7%	Multiple brands appear, no leader
No owner (demand-weighted)	89.3%	Share of estimated AI search demand sitting in unowned categories

The interpretation: in traditional search, ranking means displacing an entrenched competitor. In AI answers, most categories haven't decided who to entrench yet. The cost of entering a category now and the cost of taking it from an established owner later are not the same.

The numbers carry a caveat — the ownership threshold (4-of-5 prompts, 5pp lead) drives the 15.2% figure, and changing it moves the ratio. The safe reading is directional: owned categories are a minority.

Why the open window should worry you — 90.4% retention once it closes

The study's more important finding is not the empty 89.3% but the durability of ownership once established. Brands that became clear owners kept the top position in 90.4% of month-over-month comparisons.

"Topical authority matters in AI search, and the reason is durability. This study shows AI answers develop topic owners: Once a brand earns an outsized share of mentions in a category, it usually keeps it." — Kevin Indig, Growth Memo

Read together, the two numbers change the strategic frame. The 89.3% of unclaimed space is not permanently open — it is space that hardens the moment someone fills it. Deciding that "AI search is early, we'll respond later" means accepting that if a competitor becomes the owner first, their name gets spoken in that category next month, and the month after, with nine-in-ten odds.

Citations are the ticket, not the seat — top-cited matched top-named only 20.8% of the time

The data also breaks the assumption that stacking citations and backlinks eventually earns brand mentions. The most-cited domain matched the most-named brand in just 20.8% of categories, and the correlation between citation volume and naming frequency was -0.229 — effectively unrelated (Growth Memo).

This points the same direction as the same researcher's June study — the ghost citation phenomenon, where 62% of citations never turn into brand mentions. In AI answers, citation (your page used as a source) and naming (your brand spoken in the sentence) are separate metrics — measured separately, improved separately. Citation is the ticket into the candidate pool; naming is the seat in the answer.

Overlaid on the Korean market

This is a US-only, ChatGPT-only study, so the figures do not transfer directly. But RanketAI's measurement of a 12-company Korean B2B SaaS public-web sample points the same way — the ChatGPT mention rate was 33%, less than half of Gemini and Perplexity (83%), with ChatGPT categories comparatively emptier (RanketAI benchmark, 2026-07). Twelve companies is an early observation, not a conclusion — but there are grounds to believe the share of unowned Korean-language categories is at least as high as in the US.

Practice — three steps to claim a category

Moving from statistics to execution takes three steps.

1. Pick the categories you can win first. You cannot win every question. Define the 10–20 categories your customers actually ask about, split each into the five prompt types (definition · comparison · alternatives · use case · purchase decision), and measure the current state — which categories are open, and which already have an owner — before setting entry priorities.
2. Read leads by size. By the study's threshold, ownership requires a 5+ percentage-point lead; a 1–2pp lead is a contest that can flip any month. Even where you rank first, a narrow gap means "still competing," not "won."
3. Improve citations and naming separately. Content that earns citations (statistics, data pages) differs from assets that earn naming (comparison pages, clear positioning, third-party mentions). As the 20.8% match rate shows, doing one does not deliver the other.

Doing this by hand means categories × prompt types × repeated runs — a heavy matrix. Measurement tools such as RanketAI track brand mentions and citations separately with repeated sampling in AI Brand Visibility Analysis, and Competitor Comparison tracks the gap against competing brands on the same questions — the "lead size" in step 2.

Frequently asked questions

Our category already has a clear owner — should we give up?

No, but the strategy changes. The 90.4% retention rate means head-on contests are expensive, so the rational route is to build ownership first in adjacent open-competition categories (the 53.7%) or in sub-segment questions, rather than attacking the owner's flagship category. Measuring which adjacent categories are open is the first step.

If we keep building citations and backlinks, won't naming follow?

The data says no. The top-cited/top-named match rate is 20.8% and the correlation is -0.229. Citation gets your page into the source pool; naming is your brand appearing in the answer's sentences. Naming requires separate brand signals — comparison pages, clear positioning, third-party mentions.

Does a US study apply to the Korean market?

The direction is useful; the figures are not portable. Korean-language queries are a different sample, and there is no reason to limit measurement to ChatGPT. In an early Korean sample (12 B2B SaaS companies), ChatGPT mention rates were lower (33%) — pointing to emptier categories. Measure your own category directly.

What should we measure first to check category ownership?

Build the questions your customers actually ask, per category, across the five prompt types, then run them repeatedly across multiple AI platforms and record ① whether your brand appears ② how consistently (how many prompt types) ③ the gap to the runner-up. AI answers vary by run, so repeated measurement — not a single check — is the premise.

How many percentage points of lead is safe?

The study's ownership threshold is 5+ percentage points over the runner-up. Below roughly 3pp is safer to treat as an open contest. And before the gap, check consistency — a lead that only shows up in some prompt types is not ownership.

Related reading

• Ghost Citations — 62% of AI citations never mention your brand — the same researcher's June study; the citation-vs-mention measurement gap.
• When AI recommends your competitors but not you — diagnosing and closing the gap in categories that already have an owner.
• Korean B2B SaaS AI visibility benchmark, 12 companies — platform-by-platform mention-rate gaps in a Korean sample.