Measured data

How often do AI assistants name brands?

There is no single number. Whether an assistant names a company depends on the engine, on whether it searches the web, on the niche, and above all on whether you ask by brand name or the way a customer does — by category. Below are rates by condition, from our own measurements, each with the number of observations and an interval.

By name vs by category — the gap that matters

The single most important split: ask a model about a company by name and it usually describes it fine. Ask it the way a new customer does — by category, without the name — and the company may not appear at all. In one measured case, a company was named in 27 of 30 by-name answers but in 0 of 132 by-category answers (interval 0–2.8%). The by-category number is the one that holds your new customers.

Search mode vs memory mode

An engine that searches the web names companies far more readily than one answering from the model's memory. In a measured niche (car detailing), engines with web access named some studio in nearly 100% of answers, while a memory-only engine named one in just 17.6%. The practical takeaway: to be named, you must be present where the search step looks — the web sources, not the model's training.

Where the answer actually comes from

Across thousands of measured answers, a brand's own website appears among the cited sources only about a third of the time (32.7%), interval 31.3–34.0. Maps and directories dominate the rest — in one engine's answers, directories were in the sources of 86.6% of answers. Being named is mostly decided off your own site, on the platforms the model reads.

Provenance: these shares are measured on buyer questions in the Russian-speaking market, where almost no one else measures Alice and Yandex. The mechanism — models answer from third-party sources, with large variance between runs — is the same across markets. Open data, by niche, with intervals: /barometr.json.

No favourites: the long tail of names

Engines don't hold a stable leaderboard. In one niche, of 514 different companies that surfaced in answers, 88% appeared exactly once (interval 84.8–90.5). That cuts both ways: a small company can get named, and an established one can drop out just as easily — which is why a single snapshot proves nothing and repeated measurement with an interval is the only honest read.

How to read any 'AI mention rate' (including ours)

We don't sell a single flattering number or a growth percentage — we show your rate by condition, with the interval, and who gets named instead of you.

Measure your own rate

Free, no sign-up: your buyers' questions across several engines, the share of answers that name you by name and by category, observations and intervals — and the companies named instead of you.

Run a free check

Common questions

Is there a single "AI brand mention rate"?
No. It depends on the engine, search vs memory, the niche, and by-name vs by-category. Any single percentage without those conditions is close to meaningless, which is why we publish rates by condition, with n and an interval.
Why do different tools report very different numbers?
They measure different things — different engines, prompts and definitions of a "mention" — and most report a point estimate with no interval. A rate on 100 answers and one on 10,000 are not comparable.
Where does the AI get the brands it names?
From third-party sources far more than from the brand's own site. In our data a company's own website is among the cited sources in only about a third of answers; maps and directories dominate the rest.