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. CRM turned out to be one of the cleanest examples we have run, because of Attio.
Attio is the modern CRM that startups reach for and rave about. Well funded, well designed, exactly the tool the online startup crowd talks about. If funding and affection were what the models rewarded, Attio 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. That is 40 answers per question, 80 in total.
Version one was the plain category label:
What's the best CRM software for a business?
Version two described the buyer Attio is actually built for, with no tool named:
What should an early-stage startup use to track deals and customer relationships without a heavy enterprise sales tool?
The receipt
How many of the 10 runs named each tool, per model.
Best CRM For an early-stage startup
(category label) (no heavy enterprise tool)
Model Attio Salesforce Attio Salesforce
----------------- ----- ---------- ----- ----------
ChatGPT 0/10 10/10 0/10 0/10
Claude 0/10 10/10 2/10 7/10
Gemini 0/10 10/10 0/10 0/10
Perplexity 0/10 10/10 0/10 8/10
----------------- ----- ---------- ----- ----------
All four models 0/40 40/40 2/40 15/40
Named instead, both questions: HubSpot, Pipedrive, Zoho, and on the
startup question, Notion, Airtable, and Streak.
Attio is not phrasing-sensitive. It is just absent.
This is the part that makes CRM different from other categories we have run. With some tools the naming swings depending on how you ask. Attio does not swing. It was named zero times out of 40 on the plain question and 2 times out of 40 on the startup question. Across all 80 answers it came up twice, both on Claude. The category is owned by Salesforce, HubSpot, Pipedrive, and Zoho, and Attio is simply not in the conversation.
The models can reason about startup fit. They just do not know Attio belongs there.
Here is the tell that this is not the models failing to understand the question. Look at Salesforce on the second question. It went from 40 of 40 down to 15. The models correctly noticed that an early-stage startup that does not want a heavy enterprise tool probably should not be handed Salesforce, so they dropped it and reached for lighter options, HubSpot, Pipedrive, Notion, Airtable, Streak. The reasoning is right there. They know how to pick a leaner CRM for a smaller team. They just have never learned that Attio is one of the answers.
That is the cleanest version of a corroboration problem. It is not that the model cannot find a startup CRM. It is that the specific sentence, Attio is the modern CRM for an early-stage startup, does not exist widely enough in the places the model reads for it to reach for that name.
What this means if you are the challenger
Funding does not move the model. Product love does not move the model. A great design that your users adore in private does not become a public sentence the model can learn from. Salesforce and HubSpot are named because two decades of comparison posts, reviews, docs, and threads taught the model they are the answer. Attio has the product. It does not yet have the corpus.
The work is not a better CRM. It is getting Attio written about, by real users, as the answer to the specific jobs it wins, in the public places these models read. Not more mentions of the category, the exact sentence tied to the job.
Check your own category
Pick your category. Ask the plain label question 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 count who gets named. If you are the beloved, well-funded challenger who still does not show up, the fix is not the product.
If you want the fast version, run your own domain through the free scan and see who AI recommends in your category and who it names instead: bersyn.com free scan.
In your category, is the tool that everyone online talks about the same one AI actually names when a buyer asks?
Top comments (0)