Who AI names in each software category
A regular report on which products ChatGPT, Gemini, Perplexity and Google name when buyers ask for a tool, category by category. We're publishing the method first, so you can judge it before you see a single number.
First report coming [Date]
Why we're doing it
Buyers have moved fast. In G2's 2026 survey of B2B software buyers, 51% said they start their research with an AI chatbot more often than with Google, and 69% had picked a different vendor than planned because a chatbot recommended it (G2).
But numbers you can check on who those tools name are hard to find. Answers change every time: in one test of about 3,000 runs, the same list of brands came back twice less than once in 100 tries (Search Engine Land). Software leans on “best of” lists more than any other industry (AirOps), yet the leaderboards that exist come from vendors, like Goodie's index across 31 industries (Goodie).
We think founders deserve counts they can check: every question, every count, every range. So we'll count in the open, category by category, and show our work.
How we'll do it
The same way we count for each outshone customer, at a larger scale.
Pick the categories
We start with software categories small companies sell in, like CRMs, help desks, form builders and scheduling tools. We publish the full list and how we chose it.
Write the questions buyers ask
Plain questions in buyers' own words, like “What's the best CRM for a small sales team?”, plus the “alternatives to” and “X vs Y” questions. Every question is published with the report.
Ask each AI tool many times
Each question goes to ChatGPT, Gemini, Perplexity and Google AI Overviews many times, on different days, in a new session with no history.
Count who is named
A product counts as named when an answer includes its name. We count how often each product is named, out of all the answers, for each AI tool.
Give every count a likely range
A different set of answers would give slightly different counts, so each one comes with a likely range, and we explain how we worked it out.
Publish all of it
The questions, the counts, the dates, the method and the answers themselves, so anyone can check our work or run it again.
How many times we ask each question, and how many categories we cover: [N runs per question, N categories]. See also how we measure for each outshone customer.
What we won't do
Research that names real products has to be careful. These are our rules.
Make up or round up numbers
If we don't have a number, we don't print one.
Crown a winner from one answer
One answer is a snapshot. We only report counts across many answers, never a league table.
Call a product “invisible”
We report how often AI tools named it, and nothing more. Being missing from AI answers says nothing about the quality of a product.
Use customers' data without their OK
Research uses questions we ask ourselves. Anything from outshone customers would need their consent first.
Hide the method
If you can't check it, it isn't research. The method is on this page, before the first number.
Say what it means for revenue
We can count names in answers. We can't count the sign-ups, trials or revenue they bring, and we won't pretend to.
What it can't tell you
Every method has limits. Here are ours, up front.
- AI tools change often, sometimes from week to week. Every count is dated, and a count from one month may not hold the next.
- Wording matters. A slightly different question can get a different answer, which is why we publish every question.
- Who's asking may matter. We ask logged out, in a new session, so our answers may differ from what a signed-in buyer with a long chat history sees.
- Google AI Overviews only appear for some searches, so for some questions there will be no AI Overview to count.
Sources
Figures on this page come from these sources. We checked each on 6 Nov 2026.
G2 The answer economy, G2's research on B2B software buyers
company.g2.com Checked
Survey of 1,076 B2B software buyers, March 2026, published 15 Apr 2026.
Search Engine Land SparkToro and Gumshoe's study of how often AI recommendation lists repeat
searchengineland.com Checked
About 3,000 runs across ChatGPT, Claude and Google's AI.
AirOps Which page types earn AI citations
airops.com Checked
AirOps sells content software. It doesn't give the sample size for every figure.
Goodie AI Visibility Index
higoodie.com Checked
A vendor's leaderboard across 31 industries.
Want the first report?
It comes out [Date]. We'll post it here and in the changelog. Journalists and researchers can write to us for the data.
Or check your own product now.
Type your product's name or website. In about a minute you'll see what ChatGPT, Gemini, Perplexity and Google AI Overviews say when a buyer asks for a tool like yours.