The typical brand appears in 1 in 5 AI responses. It gets recommended in almost none.
We ran buyer-intent AI queries across 241 B2B software brands in three categories. The median brand appeared in 20 percent of responses. Its recommendation rate was zero. That gap is not unusual. It is the norm.
What we measured
For each brand, we ran 10 buyer-intent prompts (recommendation and comparison questions a real buyer would ask) across four web-grounded AI engines: ChatGPT, Claude, Gemini, and Perplexity. We counted every response where the brand was mentioned by name, and every response where the brand was actively named as a recommendation or a top choice.
The 241 brands span three software categories: project management software (173 brands), customer support software (30 brands), and HR software (38 brands). All were US companies.
Mention and recommendation are different events. Mention means the brand name appeared somewhere in the response, including in lists of alternatives, caveats, or passing comparisons. Recommendation means the brand was named as a suggested tool, a top pick, or a primary answer to the question being asked.
The numbers
| Category | Brands | Median mention rate | Completely invisible (0%) | Zero recommendation rate |
|---|---|---|---|---|
| Customer support software | 30 | 28% | 3% | 80% |
| HR software | 38 | 20% | 3% | 87% |
| Project management software | 173 | 20% | 12% | 94% |
| All categories | 241 | 20% | 9% | 91% |
What a 20% mention rate actually means
The median brand showed up in 8 of the 40 AI responses we ran. That sounds reasonable. But most of those appearances were not the outcome a marketing team wants.
In a typical "mentioned but not recommended" response, the answer named a clear winner (Asana, ClickUp, Zendesk, BambooHR), then added a sentence like "there are also tools like [brand], [brand], and [brand] depending on your needs." The brand appeared. It did not get the referral. The buyer's attention was already pointed at someone else by the time the brand's name came up.
Recommendation rate tracks the responses where the brand was the answer, not a footnote. For 91% of the brands in this study, that count was zero.
Invisible is not the main problem
The finding that surprised us: complete invisibility (0% mention) is relatively rare. Only 9% of brands across all three categories had zero mentions in any of their 40 responses. In customer support and HR software, that number dropped to 3%.
Most brands have enough presence, enough coverage on third-party sites, enough mentions in forums and comparison pages that their names show up at least occasionally when AI engines pull responses together. The problem is not that AI does not know them. The problem is that knowing a brand and recommending it are very different outcomes, and the engines are applying a much higher bar to recommendation than to mention.
Why some brands get recommended and others get listed
The brands that consistently received recommendations in this data shared a pattern in their citation sources. When a buyer asked "what project management software should I use for agencies," the engines that recommended a specific brand were pulling from sources that called that brand out as the answer: a listicle that ranked it first, a comparison post that said it was best for agencies, a forum thread where someone said "I switched to [tool] and never looked back."
The brands that appeared in passing were typically on the same sites, but as entries in a broader comparison table rather than as a recommendation. Both brands were "covered." Only one was being recommended.
The distinction is whether the coverage is presenting the brand as a conclusion or as one of many options. Engines tend to surface the conclusion.
What this means for B2B software brands
If your brand is getting mentioned in AI responses but not recommended, the problem is usually not awareness. It is authority. The sources that AI engines pull from have you on the list but not at the top of it.
The fix is different from an SEO fix. You do not need more pages. You need the pages that already exist about your category to name you as the answer, not as one of ten options. That means targeted placements in the editorial sources that engines weight heavily for your category, not blanket presence-building.
The brands in this study with recommendation rates above zero had, in most cases, one or two sources doing most of the work: a category review site, a community thread, a comparison guide that named them first. A small number of high-authority placements were producing the majority of the AI referrals. That is the lever.
We run the same analysis for your brand: 40 buyer-intent queries across four AI engines, your share-of-voice vs. your top competitors, and which specific sources are driving their recommendations. The audit tells you exactly what to target.
Get the AI Visibility Audit, $490Methodology note
All 241 brands were run on live, web-grounded AI endpoints in June 2026 using search-enabled versions of ChatGPT (OpenAI Responses API), Gemini (with grounding), Perplexity (Sonar), and Claude. Each brand received 10 buyer-intent prompts generated for its specific category. Mention and recommendation scoring was applied by the same automated scorer across all brands. Data from branded queries (searches that include the brand name) was excluded. Only organic buyer-intent results are included here.