Of the 20 companies we tracked, 99% were described correctly when we asked an AI assistant about it by brand name. When we asked the category question, the same AI named the B2B company in 35% of answers and the consumer brand in 51%.
That gap is where the breakdown happens. AI assistants know who you are. Do they cite you, or mention you, when it counts? In the category question, the one a buyer is more likely to ask before they know which brand to look for. For most B2B companies the answer is no.
For 14 days, we tracked 100 prompts across ChatGPT, Claude, Microsoft Copilot, Google AI Overviews and Perplexity. These prompts covered 20 U.S. startups, 10 B2B and 10 consumer, five prompts each. One prompt asked about the company by name. Four asked the category questions its buyers ask. That produced around 5,400 answers. Here is what our analysis showed:
Known is not the same as recommended
| Share of answers naming the company | B2B (10 companies) | Consumer (10 brands) |
|---|---|---|
| Asked about the company by name | 99% | 100% |
| Asked the category question | 35% | 51% |
| Median company, category questions | 18% | 56% |
The branded numbers are not interesting. If you ask an assistant what a specific company does, it tells you, and it is usually right. This mirrors what we would expect to see in the SERPs. However, people don’t buy that way, they usually ask for the best option, a personalized recommendation, or a comparison. When those questions are asked, most of the companies disappear from the results.
Six of the 10 B2B companies appeared in fewer than one in four category answers, and three of them in fewer than one in 10. Consumer brands did better but not uniformly: four of 10 were below one in four.
The absences were not noise. On 10 of the 80 category questions, the company did not come up in a single answer across 14 days and five engines. On 26 of 80, it appeared in fewer than one answer in 10.
In B2B, it's winner-take-most
The B2B panel has no middle. Three companies are named in 80% to 90% of category answers; the other seven average 13%. Consumer brands taper more gradually: five above 60%, one near half, four at the bottom.
The practical reading is that AI assistants settle on a short list for each category. If you are on it, you are in almost every answer. If you are not, you are close to invisible, however well the assistants know you. When a company was absent, the answers were not empty. They named other vendors.
Where companies actually start
The 20 companies above are funded, growth-stage startups. Most companies are not starting from 35%. When we run a first 40-prompt audit for a company, the picture is usually closer to zero. In the audits we ran, B2B companies appeared in only 1% to 2% of category answers, and were absent from approximately 66% of the category questions. A consumer services brand appeared in 16% and was absent from about half.
In every case the branded prompts looked fine. That is why a branded check doesn’t tell you the whole story. If you want to know where you stand, ask the questions your buyers ask.
What to do about it
- Audit the category questions, not your name. List 30 to 40 questions a buyer might ask before they know you exist. Research questions around best options, recommendations, comparisons, pricing. Check whether you show up in the answers. Leave out brand-name prompts; they score well and tell you nothing.
- Get into the sources AI assistants read. The answers that didn’t mention the brand still named other vendors, and those names traced back to comparison articles, review sites, analyst lists and community threads. Earning a place in those sources is most of the work. Your own site alone does not put you on the shortlist.
- Answer the category question on your own site. Pages built around the buyer’s problem (“best X for Y,” “X vs. Y,” “how much does X cost”) give the assistants something to cite. Your product pages written for people who already know who you are, do not.
- Measure the same way every time. Coverage is share of answers, not a simple yes or no, and it moves as the engines refresh their sources. Rerun the same prompt set monthly to measure the impact of your changes.
How we ran the study
We tracked 100 prompts across ChatGPT, Claude, Microsoft Copilot, Google AI and Perplexity from Sept. 30 to Oct. 13, 2026. We used U.S. settings, via the Otterly.AI monitoring platform.
All 20 companies are venture-backed or privately funded U.S. companies with 100 or fewer employees and estimated annual revenue between $1 million and $15 million. The 10 B2B companies were in the following industries: healthcare data, financial infrastructure, accounting AI, developer tools, employee benefits, and construction data. The 10 consumer companies were personal finance, consumer health, travel, children’s education, food and home products. None of them are current Veuno clients, and we are not naming them.
Each company had five prompts: one branded “What is [company] and what is it known for?” and four category questions with no brand named: a “best [category] in 2026” request, a recommendation request, a use-case question, and a comparison question. A company counted as appearing when its name was in the answer text. The study produced more than 5,400 answers, about 4,350 of them to category questions. Results varied by engine, and Perplexity figures rest on a smaller number of completed runs
Find out where you stand
Want to run the 40-prompt version of this for your own company? Both branded and the category questions your buyers actually ask? With a read on which sources the assistants are citing instead of you? That is what we do.
