Track how AI recommends your product

Send the questions your buyers ask, and the ones your competitors win on. Each is asked 4 ways (list request, buyer asking for advice, search query, worth paying for), because the wording alone can move a product in or out of the answer.

On the first of each month every model answers each question 3 separate times per wording, and the results are saved as a dated snapshot. A single answer can be luck; repeated runs show what each model reliably recommends.

A change since last month counts only when it is larger than two standard errors, and only on models that stayed the same between runs. A lab swapping in a new model never reads as your product rising or falling.

See each model's one-line reason for its picks, the models that never name you, the competitor each model favors, and how your share moves with the wording of the question. Each model's full answers are published on the question's own page.