We checked 174 competitor lists against 31,561 AI answers
Who does ChatGPT recommend when it doesn't recommend you? Most teams answer with the list they already have: the names in the sales deck, the ones they lose deals to. We wanted to know how well those lists match what AI engines actually tell buyers.
Every brand our users track has a competitor list, a median of eight names, almost all set when the brand was added (1,390 of 1,453 in its first hour). We took the buyer questions those brands track, the ones that don't name the brand, as asked on ChatGPT, Claude, Perplexity and Google AI Overviews from October 4 to November 7, 2026. After dropping repeated identical answers and brands with fewer than 30 answers, 31,561 answers to 3,457 questions remained, for 174 brands.
For each brand we ranked every company the answers named, other than the brand itself, and checked each one against its list.
The most-named company was on the list 87% of the time
For 151 of the 174 brands (86.8%, 80.9% to 91.0%), the company the engines named most often was already a listed competitor. People know who leads their category. That part of the list is fine.
Only 6 of the 10 most-named companies were on the list
Below the leader, the match falls away. The median brand had 6 of its 10 most-named companies on its list. 86 brands (49%) had five or fewer. 26 had three or fewer.
The missing names weren't rare. For the median brand, the most-named company it didn't list showed up in 17.7% of its buyer answers, and three unlisted companies each showed up in at least a tenth of them. Only 23 of the 174 brands had no unlisted company that common.
One answer in four named no listed competitor at all
29,405 answers named at least one company besides the brand. In 7,345 of them (25.0%, 24.5% to 25.5%), none of those companies was on the brand's list. The buyer got a shortlist. A team watching only its listed competitors would see an empty answer.
Perplexity did this most often, in 28.9% of its answers. Claude was at 25.8%, ChatGPT at 24.0% and AI Overviews at 21.5%.
20 of the 30 unlisted names we read were real competitors
An unlisted name isn't automatically a competitor, so we read the top unlisted company for 30 random brands by hand. 20 were direct competitors the brand hadn't listed. 3 were another name for a listed company, like an acronym or a former name. 7 weren't competitors: an editing app in answers about a photography course, a coding agent in answers about developer tools, a certification body in answers about exporting produce.
One example, without naming the brand: a bookmark manager listed four competitors. Across its 269 answers, three read-later apps it hadn't listed were named in 36%, 30% and 27% of them. That's more often than three of the four it had.
One listed competitor in ten was never named
It runs the other way too. 148 of the 1,453 listed competitors (10.2%) weren't named in a single buyer answer for their brand in five weeks. They may be real rivals in a sales call. In these answers, they weren't on the shortlist.
What we'd do
- Build the list from the answers. Rank every company named across your buyer questions and look at the top ten. Expect about four you don't track.
- Read the answers that named each new company, and the pages they cited, to see how it got there before you decide what to do about it.
- Keep listed names that never appear if they matter in sales, but don't let them set your benchmark for AI answers.
How Proofsource handles it
Proofsource records every company named in the answers it collects, including ones you don't track. When an unlisted company keeps coming up, the Today page shows it with a Track it button, and adding it fills in its mentions from answers already stored.
Where this could be wrong
Names come from our extraction, and matching a name to a list entry is loose: we counted a match on the domain or when one name contained the other. That still misses acronyms and renamed companies, which turned up in 3 of the 30 we read, so the true overlap is a little higher than we report.
Most lists were suggested by Proofsource at signup from a read of the brand's site, then accepted or edited by the user. A list built by hand from lost deals might match better, or worse.
The 174 brands are the ones our users chose to track, many small and many in India. This is one month, four engines and buyer questions that don't name the brand.
