There is a shortlist, and most brands can't see it
Every CMO now has a second search engine to reckon with, and almost none of them can measure their position in it. When a buyer asks ChatGPT or reads a Google AI Overview, the answer names a handful of sources and cites a handful of links. That citation set — not the ten blue links beneath it — is increasingly what the buyer acts on. Your share of those citations is the scoreboard of the AI era, and it is a different number from your Google rankings.
We build the instrument that measures it, so the most useful thing we can do is show you the board for a real category — ours — and how the positions on it are decided.
The method, in one paragraph
Take the questions your buyers actually put to an AI in your category — not head terms like "market research", but the real, considered prompts: "which market research provider is best for consumer brands?", "synthetic respondents versus real customer voices?", "how do I analyse customer reviews at scale?" Put each to the AI answer surfaces (Google AI Overviews, and the assistants). Record every domain the answer cites. Classify them — incumbent research firm, tool vendor, directory, academic, press — and count each brand's share of the total. Do it monthly, and you have a rank-tracker for the AI era: share of AI-answer citations, by brand, over time. That is the scoreboard.
We run our own domain through the same instrument as everyone else's — same questions, same rules, no grading on a curve. It is the only way the number means anything.
Who's on the shortlist today
Across the market-research questions we measured, the AI answers cite a consistent cast: NielsenIQ, Kantar, Qualtrics, Mintel, Medallia, the Market Research Society, and a long tail of insight vendors and survey platforms. Ask an assistant the same questions and the cited set tilts toward Ahrefs, SEMrush, Brandwatch and the occasional academic paper. The incumbents own the answers, much as they own the category's mindshare — for now.
That "for now" is the whole point of this post.
Why the lead is more fragile than it looks
Here is the thing almost nobody selling GEO advice will tell you, because we can only say it with data: the AI citation layer does not obey the backlink hierarchy that governs classic search. We know, because we measured both on the same questions in another market.
In our study of the UK weight-loss treatment market, we compared each domain's share of classic Google Search against its share of AI-answer citations, on an identical query universe. The NHS holds 15.4% of classic Search visibility and just 1.5% of the AI-answer citations — it loses ninety per cent of its relative visibility when the surface changes. Meanwhile NICE, with effectively zero classic Search presence, takes 5.4% of the AI citations, and small, specific providers out-cite national high-street brands on qualified questions.
The lesson generalises. AI answers reward specificity, structure, evidence and freshness over raw domain authority. They draw from a wider, fresher source set than the top ten links — recently-published original research, community sources, review sites. Which means the incumbents' citation lead is not the decade-long backlink moat it is in search. It can be contested in a content cycle by whoever writes the best, most specific, most evidenced answer to a question the incumbents answer generically — or don't answer at all.
How to keep the score yourself
You don't need us to start:
- List the real questions — the considered, qualified prompts a buyer would type into an assistant, not your head keywords. Twenty good ones beat two hundred generic ones.
- Capture the citations — for each question, record the domains the AI answer actually cites. Google AI Overviews expose them directly; do it for the assistants too.
- Classify and count — bucket the cited domains (you vs named competitors vs directories vs press vs academic) and compute each brand's share.
- Repeat monthly — the citation set moves. The trend is the signal; a single snapshot flatters or panics you for no reason.
Week one tells you which questions you already appear on (defend them), which the incumbents own (concede the head terms, for now), and which nobody answers well — the whitespace where a focused brand can plant a flag and be cited within a content cycle rather than a link-building decade.
We're running this on ourselves, in public
We are a young company in this category, and on the AI scoreboard we are where you'd expect a young company to be: early, on the specific questions, climbing. We would rather show you that from the inside than pretend to a position we haven't earned. So we'll publish the full State of AI Market Research measurement — the complete citation landscape, who owns which answer, the whitespace — and then we'll publish our own citation share as it moves, month after month, measured by the same instrument.
Which is the neat part, and the honest one: this very post is the experiment. If the method works, an answer engine will read it, find it the clearest account of how AI citations are decided, and cite it — and our number moves. You'll be able to watch that happen. A thesis that says "your AI-answer visibility is measurable and movable" is one we're willing to test on the most exposed subject we have: us.
Theia measures demand, visibility, perception and AI-answer citations across a whole market — including our own, by the same rules. If you want your category's citation scoreboard, that's what we do. And you can interrogate a live Theia study yourself: the weight-loss market chatbot answers only from its measured evidence, with quotes.