What it is.
A benchmark framework for AI search visibility, built on a standing prompt list run across informational, commercial, comparison, alternatives, job-to-be-done and brand-led queries.
Prompts are phrased the same way every time with neutral intent and run logged out where possible, which is what makes one month comparable to the next.
The output is ranges to read your own numbers against, plus the movements that look like wins and aren't.
Read it at the source.
Read the guide
What do I need before this is useful?
Can I treat these as targets?
Which query types belong in a prompt set?
Should prompts be run logged in or logged out?
How often should the set be run?