Statistics
AI search statistics, counted from our own corpus
Between June and August 2026 we ran buying questions through ChatGPT and Perplexity every week for the brands we tracked and kept every answer. These AI search statistics come out of that dataset: which pages get cited, whether those pages rank for the same question, and how the engines differ. Nothing here is a re-list of someone else’s study, and every figure carries its sample size.
- 53%
- 115 of 218 cited page-and-question pairs, 203 distinct pages
- 20%
- 43 of 218 cited page-and-question pairs
- 28%
- 60 of 218 cited page-and-question pairs
- 10 vs 3
- 362 Perplexity answers and 104 ChatGPT answers
Reports
Each one is a single run, dated, with its method and its limits on the page.
Questions
Where do these AI search statistics come from?
From the answers CueScout logged for the brands it tracked. Each report says which engines, which dates and how many answers, and every figure carries the count it was computed over.
Why is the ranking index Brave and not Google?
Brave runs its own index rather than reselling Google’s, and it is the index we can query 100 results deep through an API. The reports say so next to every rank figure.
Can I quote these numbers?
Yes, with the sample size attached. Each report ends with the attribution line we would like you to use.
Run the same checks on your own brand
One buying question, one engine, free and with no signup.