What gets a brand named by ChatGPT, Gemini and Perplexity, measured from 21,534 answers to real buying questions in niche markets.
of B2B software buyers start research in an AI chatbot more often than in Google
say a chatbot led them to a different vendor than they planned
bought from a vendor they'd never heard of before
The question is no longer “do we rank?” It is “does the answer name us?”
When an engine didn't use a brand's own site as a source, it almost never named the brand. When it did, the brand was named in a fifth to most of the answers.
Without their own site among the sources, most brands were named in 2% of answers or fewer.
Perplexity used brand sites in over half its answers but named the brand in fewer than one in four.
Competitors' sites appeared in 70% of answers. Wikipedia appeared in 8 out of almost 18,000.
When a buying question named a competitor, the client was named in just 0.3% of answers.
All three AIs name you together only 16% of the time. Track how often you're named over weeks.
See the numbers behind each finding in the full report
Read the reportQuestions that mention a place named the client 2 to 6 times as often. Local specialists win where the field is smaller.
Most clientsSmaller brands got named first on narrow questions only they answer well. Generic category questions come later.
Smaller clientsSite in the source list but not quoted in the answer: named 4%. Quoted: 50%.
Every clienttypical wait from publishing to first use, for posts AI picked up
of answers that used a client's site used one of our articles
So what: Publish steadily — when AI picks up a new post, it usually does so within a week.
Five findings, three more patterns, a six-step playbook and the method behind every number.
Say who it's for, the price and what's different, in plain text. Don't block the AI search crawlers.
Strong · Findings 1, 2Find the "best X" lists, comparison pages, review profiles and Reddit threads cited for your questions. Get onto them.
Moderate · outside researchOn how-to pages, say plainly where your product fits, with a comparison table that includes you.
Moderate · Finding 2"[Incumbent] alternatives" and "[Incumbent] vs [You]", with honest tables. Aim for third-party alternatives lists too.
Moderate · Finding 4Prices, timelines and data, with sources. Publish steadily; posts AI used were picked up in about a week.
Mixed evidenceOver weeks, not one answer. Build your competitor list from who the engines actually name.
Strong · Finding 5In Ahrefs' study of 75,000 brands, link-based authority mattered far less than brand mentions on other sites.
Adding schema didn't change ChatGPT citations in Ahrefs' test of 1,885 pages. llms.txt showed no link across 300,000 sites.
Just 8 of almost 18,000 answers to niche buying questions used Wikipedia.
The complete deck as a PDF, with every chart, sample size and source. We'll also send you the Q4 edition when it's out.
We'll run the same analysis on your category and show you where you're read, where you're named, and who gets recommended instead.