In most categories, being named is not the hard part
We counted which brands 160 AI answers named across eight software categories. Two of 67 were named in none of them, and 63% appeared in at least half — which is not what this category sells.
Last updated 23 September 2026
What we counted
We run a public index of eight software categories. For each one we take ten buyer questions, ask every question of DeepSeek and Kimi, and count which brands come back. That is 160 answers, and 67 brands in total.
Each question was asked once per engine. That matters, and we come back to it below.
What came back
Mention rate is the share of a brand’s 20 category answers that named it. Across all 67 brands:
- 2 of 67 — named in none of their category’s 20 answers.
- 5 of 67 — named in fewer than one in ten.
- 42 of 67 — named in at least half.
- 18 of 67 — named in three quarters or more.
Two brands were named in every single answer we asked about their category: Zendesk in customer support, and PostHog in product analytics. In our own category, AI visibility tools, Profound led at 17 of 20.
So the picture is not the one this category usually paints. The bottom is nearly empty: only five of the 67 are named in fewer than one in ten answers, while 42 sit at half or more. A handful of brands are close to invisible, and far more are named routinely.
The caveat that matters more than the finding
These are single draws. Each question was asked once per engine, so a brand at 85% is not a brand that is recommended 85% of the time — it is a brand that came back in 17 of 20 single readings. Our own rechecking, in an earlier post, found 98% of questions gave the same answer every time we asked; that is why we now hold a third of our questions back as a control group. Still: read the shape, not the individual numbers. The shape survives the noise. A particular brand’s 85% does not.
Three more limits, stated plainly:
- Sixty-seven brands is the list we assembled for these eight categories, not every brand competing in them.
- Two engines, English questions, one reading each.
- Eight categories is not a market. It is a sample we can afford to re-run, which is a different and more useful property.
Why we are publishing a finding against our own pitch
The pitch in this category — ours included — is "AI answers do not mention you; here is how to fix that". These numbers do not support the first half. In a category of eight or nine brands, about two thirds of them are named in half or more of the answers, so being named is close to a baseline rather than an achievement.
What is scarce is share. The leading brand in each of the eight categories took between 15% and 26% of the mentions — nobody is the answer. That is the number worth buying and the number worth checking: not "am I mentioned", but how often, and ahead of whom.
We would rather publish that than a number we can only get by being vague about the sample. Every figure here comes from files you can read: the category pages list the questions, the engines and every brand’s counts, and the free check runs the same measurement against your own domain, whether or not you are in our index.