What our '144 companies' stat actually measures
You'll see us cite a number like this: half of the B2B brands we tested were invisible or misidentified on ChatGPT. It's a good stat. It's also easy to misread, so here's the full methodology — what we actually measured, and where the number stops being fair to generalize.
What we actually tested
We picked 144 B2B SaaS companies across ten niches — HR software, sales tech, DevTools, fintech, and so on. But we didn't pick them at random. We specifically selected companies with generic or ambiguous names: dictionary words (Monday, Copper, Remote, Mercury, Current), geographic terms, or names that collide with a bigger, more famous entity (Apollo, Dagger, Fabric).
For each one, we ran a single prompt with no domain, no context — just the name:
"What is {company}?"
Then a judge pass classified every response as correct (identified the right company), partial (mentioned the company, but not as the primary answer), or failed (described the dictionary word, listed unrelated possibilities, or named a different company entirely).
The numbers
| Engine | Correct | Partial | Failed outright | Failure rate | |---|---|---|---|---| | GPT-5.4 | 38 | 33 | 73 | 50.7% | | Claude Sonnet 4.6 | 139 | 0 | 5 | 3.5% |
Seventy-three of 144 companies got a GPT response that had nothing to do with their actual product. Monday became "a day of the week." Copper became "a chemical element." Remote became an adjective. Apollo became the GraphQL library. Claude, tested on the exact same list, got 96.5% right.
That 50.7% is where "half were invisible or misidentified" comes from.
What this doesn't prove
This is not a claim that half of all B2B brands are invisible on ChatGPT. It's a claim about a specific, deliberately adversarial slice: companies whose names are also common words, places, or other products. If your company is named something unambiguous, your odds of a correct answer are much better than 50/50 — Claude's near-perfect score on this same list shows the underlying models are capable of getting it right when the name isn't fighting against a more common meaning.
The honest version of the stat is: if your B2B SaaS company has a generic or ambiguous name, there's roughly a coin-flip chance ChatGPT has no idea what you actually do.
Why that's still a big deal
A coin-flip chance is still enormous when the stakes are "a buyer asks ChatGPT about you by name, mid-evaluation, and gets told you're a plant or a physics concept." And generic names are extremely common in B2B SaaS — one-word, approachable, brandable names are a whole category of startup naming strategy. If yours is one of them, this isn't a hypothetical.
We can tell you in one afternoon whether your name is in the danger zone, and show you exactly what ChatGPT says when someone asks about you cold.