I measured if AI recommends challenger tools or just the same two incumbents. Some categories,winnable, some aren't, what does it change?
I've been building a tool that measures how ChatGPT, Claude, Perplexity and Gemini recommend software, and I ran an experiment this week that I think matters if you're trying to reach buyers who now ask an AI before they ever Google.
The question I wanted to answer: in a given category, can a challenger actually get named by AI, or does it just list the same incumbents every time?
I took three categories, asked one plain buyer question in each, ran it 5 times on each of the four models (20 answers per category), and counted who got named.
Product analytics was locked. Mixpanel and Amplitude got named 20 out of 20, every model every run. Heap 19, PostHog 15, then a cliff. If you're a new analytics tool, the top two slots are basically concrete right now.
Feature flags was wide open. LaunchDarkly led at 20 out of 20, but then Flagsmith 17, Optimizely 17, Split 15, GrowthBook 13, Unleash 13, Statsig 11. Eight tools with real, repeated share. A challenger here genuinely gets named.
Error monitoring was a mix. Sentry owned it at 20 out of 20, but the tail was deep. The interesting bit: Datadog only showed up 6 out of 20, because I asked "for a small dev team". Ask it generically and Datadog is everywhere. Add the team size and it drops out and the leaner tools surface.
The takeaway that changed how I think about my own go-to-market: the shape of your category tells you what to even attempt.
If you're in a frozen top-two category, don't burn energy trying to get named next to the incumbents on the generic question. You won't. Go win a sub-intent where the lock breaks, "for small teams", "cheapest", "for [specific use case]", the way Datadog evaporates on "small dev team".
If you're in an open category, you can just get named by describing what you actually do clearly, because the models are still deciding.
I'm turning this into a tool (Bersyn) because I kept needing to know, for my own stuff and for founders I talk to, whether a category was even worth fighting for in AI answers and where a company actually stands. If you want to see what the models say about your own category or domain, you can run it here: https://bersyn.com/?utm_source=indiehackers&utm_medium=community&utm_campaign=frozen-vs-open
Genuinely curious what shape your category is in. Drop it in the comments and I'll tell you what I'd guess before you even run it.
One implication I hadn't considered is that category maturity changes what "good positioning" looks like.
In an open category, clarity helps AI figure out where you belong. In a locked category, clarity alone may not be enough—you first need a reason for the model to treat you as the obvious answer to a narrower problem before you can compete on the broader one.
Yeah, that's the sharper version of it. In a locked category, clarity gets you correctly filed but not chosen, because the top two already own "the obvious answer." So the move isn't more clarity on the broad term, it's becoming the obvious answer to a narrower problem the incumbents don't specifically address, and letting that be your way in. It's exactly what the Datadog "small dev team" drop shows: the lock breaks at the sub-problem, not the category.
Just checking in to see if you had a chance to look at the email I sent a couple of days ago about our discussion here.
Whenever you have a few minutes, I'd be interested to hear your thoughts.
Thanks for the nudge Aryan, and sorry for the delay. I just replied to your email properly. Short version, the core of Beryxa is exactly what I meant, and I held to what I said last time about where my focus needs to be right now. Congrats on shipping it.