Real shopping questions, asked in ChatGPT, logged out, with a dated screenshot of every run. Every test shows the exact question, the engine, the date and what it named. The tests so far come from Who’s at the table?, my Thanksgiving series. Client before-and-afters join them in January.
How to read a test: a single answer is a snapshot, and answers change from run to run. Tests with three or more runs show which brands keep their seat. Your audit runs your questions three times on four engines.
Repeated test · ChatGPT, logged out · November 11, 2026 · 4 runsSix brands named at 7:18 AM. Two hours later, only Williams Sonoma kept a seat, with a different product. The Ladolea set, sold by Williams Sonoma, was the top pick in three of four runs.
Every run, in full
Research snapshot · ChatGPT, logged out · November 13, 2026 · 2 runsAt 7:24 AM a Williams Sonoma crate was the top pick and Boarderie was third. By 9:55 AM Boarderie was first, with a board named for Thanksgiving. Only 2 of 9 brands were named in both runs.
Every run, in full
Research snapshot · ChatGPT, logged out · November 12, 2026 · 2 runsPureBites was the top pick at 7:45 AM and 10:15 AM, with a different bag and price each time. Pupford went from two spots to none. Chewy was the seller on four of five cards in both runs.
Every run, in fullA competitive gap: AI has better material about them for that question.
A discoverability problem: the pages AI reads don't make the brand's case.
Each engine reads different sources, so each one needs its own fix.
A third-party source AI trusts in that category, and a place your brand should appear.
Free, with no obligation. I pick the brands whose questions make the most useful test, so a feature isn't guaranteed. If yours is picked, you see the result before it goes up.
Want your full picture now? The audit runs 30 of your buyers' questions, three times each, on four engines.