Centium Intelligence
research on how ai recommends brands
Centium measures how the major AI models search, source, and recommend across hundreds of brands. Intelligence is where we publish what that data shows: the trends shaping AI visibility, drawn from the platform itself.
Who AI Reads
A quarter of what AI cites has a person’s name on it. We read Influential Voices across 270+ brands to find out who those writers are, what they write, and where they publish: a dozen people carry half the answer on a typical brand, and most of them have never named the brand they are shaping it for.
Brand Search vs Category Search
The same AI model behaves differently depending on whether you ask it to recommend within a category or ask it about one brand by name. We compared both across 110+ brands and found it switches from salesperson to investigator: what it concludes, how it searches, and who it cites all change.
Query Fan-Out
When you ask an AI model to recommend a product, it runs its own searches first. We analyzed more than 300,000 of them to see what AI looks for, and what it cites, when it decides which brands to name.
Every report is built from the same data Centium tracks for individual brands, aggregated across the portfolio and reported without naming individual clients. The findings are meant to be cited; figures and methodology are stated in each report.
Subscribe
Get the next report first
Centium Intelligence: research on how AI models search, source and recommend brands, a few times a quarter. Unsubscribe any time.