We asked ChatGPT for software recommendations 0 times. Here is who it named.

Everyone assumes the best-known software brands own AI search. We decided to check. We took 0 SaaS companies that most people in tech could name from memory, asked ChatGPT the questions their buyers actually ask, and counted how often each brand got named. The results were not evenly distributed.

What we tested

Each company got 25 questions, 0 in total. The important detail is what we did not ask. We never mentioned the brand by name in a discovery question. Asking “is Notion any good?” guarantees ChatGPT says the word Notion, which measures nothing. Instead we asked things like “best no-code database tools” and “which CRM should I use”, then checked whether the brand showed up on its own.

For the head-to-head numbers below, both the brand and its competitors are counted over the same 11 neutral category questions, so nobody gets credit for a question that named them in the first place.

The results

Being loved is not the same as being visible

Why the leaders lead

0 of the 0 brands scored in the strong-to-dominant range. They share something that has little to do with product quality: they are written about constantly, by other people, in the listicles, comparison posts, documentation, forum threads and review sites that models train on. When a model has read a thousand “best tools for X” posts and a brand appears in most of them, that brand becomes the default answer.

The laggards tend to grow through channels models cannot see: word of mouth, private communities, sales conversations, and a beautiful product that people show each other in person. That builds a brand. It does not build a corpus.

What this means for your company

If brands this well known can be invisible in AI answers, brand recognition alone will not carry a smaller company. AI visibility is a separate thing from awareness, from traffic, and from Google rankings, and it has to be measured separately.

The uncomfortable part is that you cannot see this in your analytics. There is no referrer, no impression count, and no ranking position for an answer that never mentioned you. The only way to know is to ask the models directly and count.

Method and limits

Scores weight how often a brand appeared (50%), how high it ranked in the answer (30%), and how positively it was described (20%). Brand matching is whole-word, so “Linear” is not credited for “linearly”, and an answer stating it has never heard of a brand does not count as a mention.

Two honest caveats. This run tested ChatGPT only, so it is not a claim about every AI assistant. And model outputs vary between runs, so treat individual scores as a reading with a margin rather than a fixed number. The gap between a brand named every time and one named never is far larger than that margin.