BrandPulse
Four assistants. One brand.
Four different answers.
BrandPulse asks a fixed set of category questions across ChatGPT, Claude, Gemini and Perplexity, splits the branded questions from the unbranded ones, and reports where you are named, where you are not, and which sources are named instead. Per assistant. Per intent. Never only as a blend.
By Sri · Last reviewed: September 2026
From one run, September 2026
Three things a single number would have hidden.
The same brand scored 73% on one assistant and 18% on another
Same question set, same week. Perplexity named the brand in 8 of 11 answers. Gemini in 7 of 10. ChatGPT in 5 of 10. Claude in 2 of 11. That is a 55-point spread, and a single blended visibility number deletes it. Whichever assistant your buyers actually use, the blended score is wrong for them.
Branded questions inflate the headline by 17 points
Ask an assistant about a brand by name and it will usually mention the brand. Across all 42 questions the mention score was 62. Across the 30 unbranded questions, the ones where someone describes a need without knowing who solves it, it was 45. The gap is produced entirely by the question set. The unbranded number is the one attached to acquisition; the branded one measures recall of a name the person already typed.
Well documented, not recommended
Presence split hard by intent. Named in 8 of 12 attribute questions, the ones asking what a product is made of or how it behaves. Named in 2 of 8 recommendation questions, the ones asking what to buy. The assistants know the brand's facts and do not reach for it when someone wants a pick. That is a diagnosis, not a content gap, and it points at a different fix.
The full run is written up as a case study: one brand, four assistants, 42 answers.
What you receive
The report, and what makes it defensible.
- —A fixed question set for your category, split branded and unbranded, asked across ChatGPT, Claude, Gemini and Perplexity.
- —Presence rate, mean position and share of voice, reported per assistant and per intent, never only as a blend.
- —The verbatim sentences each assistant wrote about you, with every cited source in the order it was used.
- —A stability figure from repeat sampling, because a single run is a draw rather than a measurement.
- —Batch-over-batch comparison on the same question set when you run it again, so movement is real and not an artefact of asking differently.
Questions we get
Why not just report one visibility score? +
Because it is wrong for everyone. In the run above the blended score was 51 while the assistant your buyers most likely use put the brand anywhere between 18 and 73. A blend is an average of four different realities. We report all four and let you see which one matters for your customers.
What is the difference between a branded and an unbranded question? +
A branded question names you. An unbranded question describes the need. "Is [brand] any good" is branded. "What should I buy for [problem]" is unbranded. Only the second kind has anything to do with winning a customer who does not know you yet, which is why we score it separately and quote it first.
Can you compare this month to last month? +
Only on an identical question set. If the set changes between runs, the number changes for reasons that have nothing to do with the assistants. We hold the set fixed and compare batch to batch. We will not publish a trend line across runs that used different questions, and we will tell you when a comparison is not valid.
Do you name our competitors in the report? +
Yes, to you. Share of voice is only useful if you can see who is taking it. Nothing in the report is published anywhere; it is delivered to you and stays yours.
See what AI says about you.
A BrandPulse run takes about a week from question set to report. Start with the free check to see one assistant's answer today, or book a call and we will scope the full run.