This is original measured research — the same instruments we point at any market, pointed at our own. All figures measured July 2026; method and reproducibility notes at the end.
Why we measured our own category
When you sell market measurement, the most credible research you can publish is a measurement of the market you sell into — with yourself in the frame, scored by the same rules as everyone else. So here it is: the AI-market-research category, read the Theia way. 37,120 evidenced observations, the full UK and US search landscapes (1,800 and 3,147 keywords; 133 and 108 ranking domains), and the citation layer — what the AI answers actually recommend when buyers ask.
Five findings. Three of them surprised us.
1. Nobody owns the AI answers — not even the giants
Ask the AI surfaces the questions buyers actually ask — "which market research provider is best for consumer brands?", "synthetic respondents versus real customer voices?" — and classify every source the answers cite. The expectation: Gartner, Kantar and NielsenIQ own the answers the way they own the mindshare.
The measurement: the citation layer is dominated by listicles ("top 10 market research companies…"), directories, and one live methodological debate — synthetic respondents — with the incumbents appearing only occasionally. NielsenIQ leads with a handful of citations; Kantar, Qualtrics and Mintel appear; but no brand holds anything resembling ownership. The single most cited type of source is the third-party list, and the single hottest topic is a method argument the industry hasn't settled.
The implication cuts both ways. For incumbents: two decades of brand equity have not transferred to the answer layer. For challengers: the most important shelf in the category is still unstocked.
2. This category is bought on generic terms — which means content can win it
Across the combined UK+US core vocabulary, 97% of the category's search terms are generic ("market research tools", "how to analyse customer reviews") and only 3% carry any brand. Compare that with consumer categories we measure: in UK weight-loss treatment, product names carry a third of all demand.
A category bought on generic terms is a category where the language is still unowned — no player has done to "market research" what Mounjaro did to weight loss. Visibility here is won by answering the generic questions better, not by defending a name. That is unusually good news for anyone with something genuinely new to say, and unusually dangerous for anyone relying on brand gravity.
3. What buyers actually ask: education first, then five jobs
The demand structure, from the measured vocabulary: roughly half of all category search is general education — what AI market research is, how it differs from traditional research, what it costs. The other half splits into jobs: competitive intelligence (11%), SEO/keyword tooling (7%), sentiment and review analysis (7%), social listening and share-of-voice (6%) — and, already at 7% and the fastest-emerging theme: AI research and GEO — how brands appear in AI answers.
Read that again: measuring AI-answer visibility is already as large a search theme as sentiment analysis, a category that has existed for fifteen years. The market is asking for the instrument almost as fast as it is being built.
4. In B2B, the citation game is in the assistants, not Google's AI Overviews
Across our ~117 SERP-collected category queries, Google showed an AI Overview on roughly 5% — against far higher trigger rates in the consumer categories we measure. The buyer researching market-research tooling meets AI answers mostly in ChatGPT-style assistants, where the cited set skews toward SEO-tool content (Ahrefs, SEMrush), professional bodies (the MRS) and the occasional academic paper.
For B2B categories, this changes the GEO playbook: optimising for Google's AI Overviews is optimising for a surface that barely fires. The assistants — with their appetite for structured comparisons, methodology pieces and research — are the layer that decides.
5. The search landscape rewards publishing muscle, not research pedigree
The domains that dominate the category's classic search visibility are not the research giants. They are the content-marketing machines and directories: Hootsuite, SpyFu, Moz, Qualtrics' content arm, Greenbook, marketresearch.com, Investopedia. Gartner appears — but as one publisher among many, out-published by tool vendors' blogs.
The pattern matches finding 1: in this category, visibility belongs to whoever publishes the most useful answers, not whoever has the deepest research bench. The pedigree players ceded the open web to content teams; the AI layer is now repeating the same contest with the shelf still open.
Where Theia sits in its own measurement
In the frame, by the same rules: present in the ranked search set on a handful of specialist queries, cited in the AI layer on the synthetic- respondents debate — early, on the specific questions, climbing. We publish this position because we'll keep publishing it: our citation share, measured monthly with the same instrument, in public. The thesis this report rests on — that the answer layer is measurable, unowned, and winnable on evidence — is one we are testing on ourselves, and you're reading part of the experiment.
Method & reproducibility
Discovery: 39 seed terms + 20 buyer questions expanded through SERP and ranked-keyword analysis to the UK (1,800 keywords) and US (3,147) category vocabularies; 585/575 ranking domains mapped per country. Evidence: 37,120 observations enriched from articles, video, forums and AI answers. Citation layer: AI-answer sources captured per query (Google AI Overviews where present; assistant answers via prompt battery), every cited domain classified (incumbent / tool vendor / directory-listicle / professional body / academic / debate participant). Demand themes classified over the core vocabulary. The same pipeline, guards and groundedness rules as every Theia client study — each claim above traces to the corpus.
The full citation leaderboard, domain landscape and question universe are available as the study's data tables — and the follow-up to this report is a monthly scoreboard. If you want your category measured the same way: that's what we do.