- Case study
- Cognos
- Tax resolution
AI named our client in nearly half its answers.More than twice the next firm.
Our client is a tax resolution firm that runs on Cognos. On September 22, 2026 we asked ChatGPT and Perplexity the same 15 buyer questions that never mention the firm, 30 answers in all. The firm was named in 14, more than any other firm.
By Andrew Bethel · Data from Sep 22, 2026
- Unprimed answers naming the firm
- 14 of 30
- ChatGPT
- 5 of 15
- Perplexity
- 9 of 15
- Share of voice
- 42.6% · 1st of 20
From the Cognos export. One run, two assistants.
Sep 22, 2026
47%
of unprimed AI answers named the firm
14 of 30 · ChatGPT and Perplexity
Sep 22, 2026
42.6%
share of voice, 1st of 20 firms
Next firm 18.2%
Sep 22, 2026
67%
of unprimed answers cited the firm as a source
20 of 30
AI visibility measures from one run. This page doesn't report calls, cases or revenue.
Questions that never say its name.
When our questions named the firm, the assistants talked about it in 9 of 10 answers. That's expected, and it says very little. The number that matters comes from the other kind of question.
Fifteen of the 20 questions in this run don't mention the firm at all. They're what someone with an IRS problem actually types: how to stop a wage garnishment, who handles offer in compromise cases, what to do about unpaid payroll taxes, how to get penalties reduced. Each went to ChatGPT and Perplexity, which makes 30 answers.
Perplexity named it more often than ChatGPT.
Perplexity named the firm in 9 of 15 unprimed answers. ChatGPT named it in 5 of 15.
Unprimed questions only · Cognos run, Sep 22, 2026
Show data table
| Assistant | Named | Answers | Rate |
|---|---|---|---|
| ChatGPT | 5 | 15 | 33.3% |
| Perplexity | 9 | 15 | 60.0% |
| Both | 14 | 30 | 47% |
The two assistants don't agree on who to recommend, so we track them separately rather than averaging the gap away.
About these numbers.
How is this measured?
We asked the questions this firm's buyers ask and recorded which firms each answer named, and whether the firm was cited as a source. The headline rate counts only the 15 questions that don't mention the firm. The five that do are left out, because they would inflate it.
Is this a before and after?
No. It's a single run from September 22, 2026. We haven't shown earlier numbers because the run we verified this page against is the only one in the export.
Does this show more clients?
No. These are AI visibility numbers. They show how often the assistants recommend the firm, not how many people called it.
Why are there no answer screenshots?
This page reports the counts, not the answers themselves, so there are no screenshots or transcripts here.
Find out what AI says when your buyers ask.
We respect our clients' privacy, so we don't name them here. We're happy to talk through these results and share more examples on a call. The fastest way to see where you stand is your own AEO audit.