I Asked ChatGPT 100 Questions About One SaaS Category. Here's Who It Actually Recommends, and Why Your Mention Count Is Lying to You

Every project management brand wants to know if ChatGPT is talking about them. Fewer ask whether ChatGPT is actually recommending them. To find out, I dug into a teardown from High Salience that asked ChatGPT 100 real questions about project management software, then coded every answer against a fixed rulebook. The results expose a gap between showing up and being chosen, and that gap should worry anyone tracking AI search visibility using mention count alone.

The Problem With Mention Count

Most brand tracking tools count how often a name gets mentioned and call it a day. But AI search visibility isn't a popularity contest. It's a filtering process, where a language model decides which brand actually deserves the buyer's trust for a specific use case.
The High Salience data makes this painfully clear. Notion showed up in 40 of 100 ChatGPT answers but was only recommended in 15 of them. Basecamp appeared in 20 answers and was chosen in just 6. Trello was named in nearly half of all answers, yet ChatGPT only recommended it about a fifth of the time. Any dashboard that reports raw mentions would show all three brands performing reasonably well. In reality, ChatGPT was listing them for thoroughness, then sending the buyer toward Asana or ClickUp instead.

How I Tested 100 Questions

The teardown split its 100 queries into four intent types: 30 category questions ("best project management software for agencies"), 30 head-to-head comparisons ("Asana vs ClickUp"), 20 alternative-seeking questions ("cheaper alternative to Asana"), and 20 direct recommendation requests. Each query ran once through ChatGPT with web search enabled, logged out, in the U.S. English context, on September 8, 2026.
Every brand named was coded into one of three buckets: recommended for a stated use case, listed without being chosen, or mentioned only in passing. A parallel Google search was pulled for the same 100 queries to serve as a control. That pairing is what makes this data useful: it doesn't just show what ChatGPT said; it shows how differently that answer looked from a standard search results page.

What ChatGPT Actually Recommends

Asana came out on top by a wide margin, appearing in 80 answers and getting recommended in 61 of them, a 76% conversion rate from mention to pick. ClickUp followed at 69 appearances and 45 recommendations (65%), while monday.com landed at 68 appearances and 38 recommendations (56%). Jira is a standout case: it only appeared in 39 answers, but converted 67% of those into a recommendation, beating out much more frequently mentioned names like Trello and Notion.
One of the more surprising findings involved query framing. Ask about project management "for a construction company" and ChatGPT drops the entire mainstream field, recommending Procore, Buildertrend, and Fieldwire instead. Ask about a law firm, and it points straight to Clio. A single descriptive word in the prompt can swap out the whole competitive set, something no traditional keyword ranking report would ever catch.

Why Your Mention Count Is Lying To You

Here's the part that should reframe how teams think about AI search visibility: 78% of ChatGPT's recommended brand mentions went to brands whose own website didn't rank in Google's top ten organic results for that query. Google's first page for these searches was dominated by listicles and comparison sites from outlets like Forbes and Zapier. But the brand ChatGPT told buyers to actually use was still Asana.

Citations tell a similar story. Of the 362 citation events tracked across the 100 answers, only 23% pointed to a domain that also appeared in Google's top ten. ClickUp earned the most total citations at 52, largely through its own comparison and "best tools" content, even in answers where ChatGPT ultimately recommended a competitor. This lines up with independent findings elsewhere: Semrush's separate tracking of the same category found Trello with more raw mentions than Asana, but a fraction of the citations, reinforcing that mention volume and actual trust are measuring two different things.

This split matters more than ever given how quickly buyer behavior is shifting. G2's research found that 87% of B2B software buyers say AI chat tools are already changing how they research purchases, and half now start their buying journey inside a chatbot instead of a search engine, a jump of 71% in just four months. If a brand's visibility strategy still stops at "count the mentions," it's optimizing for a behavior buyers have already left behind.

The Winning Strategy For AI Visibility

Three practical shifts follow from this data. First, stop treating comparison queries as a single fight to win outright. In 28 of the 30 head-to-head queries in the teardown, ChatGPT recommended every brand named, just under different conditions. Write your own version of "choose us when X, choose a competitor when Y" before the model writes it for you.
Second, identify the vertical-specific prompts that swap out your entire category, like "for a law firm" or "for a construction company," and decide deliberately whether you want to compete for that framing. Third, stop using Google rank as a proxy for how a brand performs in AI answers. This dataset found the overlap between the two is weak enough that they need separate tracking altogether.

Mentions Won't Save Your Category Position

The teardown's clearest lesson is blunt: mentions are furniture. They fill out the room, but they don't win the sale. Brands serious about AI search visibility need to track recommendation rate, not just appearance count, and build content that states plainly who they're right for and who they're not. As more B2B buyers hand their research over to chatbots, that distinction is what will separate the brands ChatGPT recommends from the ones it just lists.