Nine of our 25 buyer questions named a tool, and the leader took 75 mentions from them
Between October 4 and 10, 2026, we ran 25 buyer questions about AI visibility tools through ChatGPT, Perplexity, Claude and Google AI Overviews on three scan days, 220 answers in all. None of the questions named us. The brand we benchmarked is Proofsource, and the leader is the tool the engines named most often, which we'll leave unnamed.
Across all 220 answers the leader was named in 130, or 59%. That's the figure most dashboards would put on the slide. But three of the questions named the leader outright, as head-to-head comparisons, and it was named in all 32 of their answers. Six more asked for alternatives to some other tool, and the leader turned up in 43 of those 52 answers (83%). The best-known name is the obvious alternative to anything.
On the 16 questions that named no tool at all, the leader was named in 55 of 136 answers: 40%, with a 95% interval of 33% to 49%. The runner-up changed too. Across every question it was a different tool, 99 answers to the leader's 130. On the neutral questions it was a broad SEO suite, 52 to 55, which is a tie at this sample size.
Questions that name you inflate your own side the same way
We also track five questions that ask about Proofsource by name. They produced 40 answers, and we were named in 30. Pool them with the buyer questions and our share of all brand mentions reads 5.7% (31 of 548) against the leader's 25.5%, a gap of about four to one. Leave them out and it's 1 mention in 500 against the leader's 130. On the 16 neutral questions we were named once in 136 answers.
A buyer who already types your name isn't the buyer you're trying to win, so neither pooled figure says much. The honest set for a head-to-head is questions that name neither side. If you need more volume, write more of those instead of borrowing questions that name a brand.
Share of voice moved from 23% to 34% when we changed who counted
Share of voice is a brand's mentions as a share of all brand mentions. The IAB's August 2026 measurement framework defines it within "a defined competitive category" and requires providers to disclose how that category is chosen and how the total mention pool is sized, because both change the result.
Here's by how much. On the 16 neutral questions, the leader held 55 of the 241 mentions among the ten tools we track, 22.8%. Two of those ten are broad SEO suites that the engines named 79 times between them. Take them off the list and the leader's share becomes 55 of 162, or 34%. Not one answer changed.
So freeze the list before the first reading, print it beside every share of voice figure, and don't edit it mid-quarter. If you must add a competitor, restate the history on the new list.
Google AI Overviews named the leader in 18 answers and cited its site once
Being named and being cited are separate measures. The IAB framework calls the second Citation Rate, the share of answers that cite the brand as a source, and it says the gap between a brand's mention rate and its citation rate is a signal in its own right. On the 16 neutral questions, 34 answers per engine, here's that gap for the leader:
- ChatGPT named it in 18 answers and cited its site in 10.
- Perplexity named it in 10 and cited it in 8.
- Claude named it in 9 and cited it in 3.
- Google AI Overviews named it in 18 and cited it in 1.
On Google, the leader's lead lives on other people's pages
Seventeen of the leader's 18 Google mentions came from answers that didn't link to its site. If you're chasing it there, study the roundups and reviews Google cited in those answers; its own pages barely feature. On ChatGPT and Perplexity its own site carries more of the weight, so the work on those engines looks different.
One counting trap: the leader's site was cited under two hostnames, with and without www, 60 and 59 times across all 220 answers. Count one and you halve its citation share. Merge every host a brand publishes on before you compare.
A benchmark you can repeat next month
Put these in place once, then rerun the same setup every period:
- Questions that name neither you nor any competitor, chosen before you look at results. Comparison and alternatives questions go in a separate column.
- A frozen competitor list, printed beside every share of voice figure.
- Three numbers per engine for you and the leader: the share of answers that name you, your share of brand mentions, and the share of answers that cite your site.
- Every hostname each brand publishes on, merged into one.
- An interval on every figure. At 34 answers per engine, a 50% reading spans roughly 34% to 66%.
Where this could be wrong
This is one category, AI visibility tools, and one brand, ours, which is new and barely named. A brand closer to the leader would see a smaller swing when the branded questions come out. Three scan days and 136 neutral answers are a small sample.
We count mentions only for the ten tools on our list, found automatically, which can miss a misspelling. A question counted as neutral if none of those ten names appeared in it, so a question naming an untracked tool stayed in the neutral set. Cited here means a linked source; the IAB also counts a brand named as a source without a link, which we didn't.
Proofsource lists every tracked question with each engine's answers, the brands they named and the sites they cited, so you can read the neutral questions on their own. Setting aside the questions that name a competitor is still a step you take yourself.
