TL;DR
- Head-to-head questions are 2.5x easier to win. Emerging GTM tools got named in 30.3% of “X vs Y” answers, and only 12.3% of “best tool for Y” answers.
- No ranking site decides these answers. ChatGPT pulled from 322 different websites, and the biggest one appeared in under a fifth of them.
- Incumbents control the category description, and a head-to-head page is how you get out from under it. Salesforce or HubSpot’s own pages appeared in 30% of answers, nearly twice as often as every review and analyst site combined. An “X vs Y” question is the one format that does not run through their definition of your market.
Ask a marketer how to get their brand into ChatGPT and you will hear the same plan. Write “best tool for X” posts. Get into the listicles. Win the category.
We ran real buyer questions for emerging GTM and sales startups through ChatGPT, twice each, and recorded who got named.
That plan targets the hardest question of the three. The easiest one is the one almost nobody writes for.
Who we looked at, and why them
We tracked ten venture-backed GTM and sales software companies: Sierra, DealHub, Affinity, Attention, Aurasell, Regie.ai, Encore AI, Linq, Day AI and Airspeed.
They sell deal desks, revenue intelligence, call analysis, CRM and prospecting tools. All of them are past seed and none of them is the default name in their category. All of them are also competing for the same buyer, which matters: a company selling into a niche nobody else in the set touches would not be comparable.
We picked this group for two reasons.
They compete against companies that have owned these categories for a decade. Salesforce, HubSpot and Gong were writing the reference pages for revenue software long before anyone asked an AI about it. If incumbency shows up anywhere in AI answers, it shows up here.
They are at the stage where the content plan is still being set. A company this size is choosing right now whether to spend the next year on category listicles or something else. That decision is worth getting right.
Nobody owns the category page any more
The first thing we expected to find was a gatekeeper. In normal search there is always one. Somebody owns the category page and everyone else fights for a place on it.
That is not what ChatGPT does.
Across the answers we collected, it pulled from 322 different websites. The single biggest one showed up in 19% of answers. Nothing else came close. So “get on the list” is not a plan, because there is no list.
The incumbents wrote the category description
If no ranking site is answering these questions, something has to be. We looked at what ChatGPT actually pulled up.
It was Salesforce and HubSpot. Not reviews of them. Their own websites.
Salesforce or HubSpot turned up in 30% of all answers. More than half of those were plain product and marketing pages: the Salesforce CPQ page, the HubSpot sales product page, Salesforce’s AI sales agent page. Another third were their support documentation.
Salesforce or HubSpot
30.0%
Only 7% of those citations came from questions that mentioned Salesforce or HubSpot. On the rest, a buyer asked a neutral question and ChatGPT reached for the incumbent’s sales page to explain the category.
That is the real problem with “best tool for X” questions. ChatGPT is not picking from a ranking. It is describing your market in your biggest competitor’s own words, and then listing who fits that description.
So the useful conclusion is an uncomfortable one: there is no play here. You cannot out-publish Salesforce on the definition of your category, and the usual advice does not survive contact with this. We checked whether partner marketplace listings help, since those pages sit on the incumbent’s own domain. They appear in 4% of answers, and on those answers the brand was named in 2 cases out of 12, against 19% everywhere else. It is not a lever.
The answer is not to fight harder for the category question. It is to go and find the question that does not depend on a category page at all.
Which is why “X vs Y” behaves so differently
A head-to-head question is a different job.
ChatGPT already has two named products. It cannot go and read a category definition, because the question is not about a category. It has to explain how these two specific things differ.
That is a page you can write about yourself and be believed on. It is also a page almost nobody writes, because the whole industry is still producing listicles.
The result is that the same brand that disappears from “best CPQ tool” shows up in “DealHub vs Salesforce CPQ”.
The gap between the winners and everyone else is enormous
The results barely moved between the two runs: they agreed on whether a brand got named 95.3% of the time, and seven of the ten companies scored identically. This is a real ranking, not noise.
The spread is what matters here. These are companies at a similar stage, selling into the same broad market, and the gap between the top and the bottom is close to total. Two of them were never named once.
That gap is not explained by funding or by how often they get written about in the tech press.
What to do on Monday
Write the head-to-head page. You against the best-known name in your category, honestly. Two and a half times the odds, and far less competition.
Stop paying for category listicles. They are the hardest question type in the set, and the reason is structural, not fixable with better copy.
Stop grading yourself on listicles. If you are not the biggest name in your category, being absent is the normal result, not a failure of your content.
Do not bother with the marketplace listing for this. We tested it. It does not move whether you get named.
How we measured it
We ran a fixed set of buyer questions against ChatGPT on 21 August 2026, for the ten companies listed above.
Questions came from each company’s own website and were split evenly between category questions, head-to-head questions and problem-solving questions. Any question that named the tracked brand was excluded, so no company could be named simply because it was asked about.
Citations are counted once per answer, so a source quoted three times in one answer counts once.
Two limits worth stating. This is one engine. ChatGPT does not behave like Perplexity or Google AI Mode, and we claim nothing about those. And while the brand results held up across both runs, the sources did not: the same question asked twice overlapped on only about a third of its citations. Which brands ChatGPT names is stable. Which pages it quotes to justify them is not.
Want this for your own category? Fireflyo runs your buyer questions through the AI engines and shows you who is answering them.
ChatGPT →
Claude →
Perplexity →