More buyers now ask ChatGPT, Claude, Perplexity, or Gemini for a tool before they ever open Google. So we keep measuring what those models actually say when someone asks for software in a category. Project management turned out to be one of the sharpest examples we have run, because of one tool: Linear.
Linear is a darling among engineers. Fast, opinionated, built for exactly the teams that talk about their tools online. If adoption and affection were what the models rewarded, Linear would be named constantly. It is not.
We asked two versions of the same question and ran each one 10 times on all four models, so a single lucky or unlucky answer could not fool us. That is 40 answers per question.
Version one was the plain category label:
What's the best project management software for a team?
Version two was the actual job, phrased the way the buyer Linear is built for would phrase it, with no tool named:
How does a fast-moving engineering team track issues and plan sprints with as little process overhead as possible?
The receipt
How many of the 10 runs named Linear, and named Jira, on each model.
Best PM software Fast engineering team
(category label) (the actual job)
Model Linear Jira Linear Jira
----------------- ------ ---- ------ ----
ChatGPT 0 / 10 10 / 10 0 / 10 10 / 10
Claude 2 / 10 10 / 10 5 / 10 9 / 10
Gemini 0 / 10 10 / 10 9 / 10 10 / 10
Perplexity 0 / 10 10 / 10 1 / 10 10 / 10
----------------- ------ ---- ------ ----
All four models 2 / 40 40 / 40 15 / 40 39 / 40
On the category label, Linear is basically invisible
Ask for the best project management software and Linear was named 2 times out of 40. On ChatGPT, Gemini, and Perplexity it was a flat zero. Jira was named on all 40 runs, and behind it the models handed back the same incumbents every time: Asana, Monday.com, Trello, ClickUp, Notion. The category label is settled, and Linear is not in the settlement, no matter how many engineering teams love it.
Reframe to Linear's actual job and it appears, but only on some models
When we asked the second question, the one that describes what Linear is actually for, Linear climbed to 15 of 40. But look at where the movement came from. Gemini went from 0 to 9. Claude went from 2 to 5. Perplexity barely moved, 0 to 1. And ChatGPT stayed at a flat zero even on the question written for Linear's exact buyer.
So the tool that engineers reach for first is a tool the biggest model will not name, even when you describe its job precisely. Meanwhile Jira held at 39 and 40. The incumbent survives both questions. The challenger only exists on the phrasing that matches its job, and only on the models willing to name it.
What this means if you are the challenger
Two things worth taking from this.
The category label is not your fight. If you sell into a category with an entrenched incumbent, the generic question is already lost, and asking it harder will not change that. Linear is proof: real adoption, real affection, still 2 of 40 on the plain question.
Adoption does not automatically become recommendation. A model names what has been written down in a form it can read, in the words the buyer uses. Engineers picking Linear in a Slack thread does not teach the model anything. The opening is the job-framed question, the specific sub-intent where you actually fit, published clearly enough and corroborated widely enough that a model reaches for you there. That is the sentence worth owning, not the category label you will lose.
Check your own category
Pick your category. Write the plain label version and the version that describes the specific job your best customers hire you for. Run both a handful of times on ChatGPT, Claude, Perplexity, and Gemini and watch who gets named as the wording changes. If you are the beloved challenger who never shows up on the generic question, you are in good company, and the fix is not a better product.
If you want the fast version, run your own domain through the free scan and see who AI recommends in your category and how the wording moves it: bersyn.com free scan.
In your category, does the tool everyone actually uses match the tool AI actually names, or are those two different lists?
Top comments (0)