Free original research
The measured state of AI search
We publish what ChatGPT, Perplexity, Gemini, Claude and Google AI actually recommend — ranked by share of model, measured reproducibly (10+ runs per engine, Wilson 95% CI), with the data and method open.
Report
State of AI Search
The headline finding: engines disagree on the "best tool," and ~95% of the citations behind their answers are third-party. What that means for AI visibility.
LeaderboardsAI Visibility Index
Per-category share-of-model leaderboards (CI/CD, observability, and more), refreshed monthly. Datasets published CC BY 4.0.
Method
How we measure
Every buyer prompt runs 10+ times per engine; we report the median share of model with a 95% confidence interval. Consumer apps and APIs differ — we say so.
Honesty. A single AI screenshot is one sample from a distribution. We never publish a figure we have not measured this way, and figures marked as a methodology preview are illustrative until the first real public run.
Want your own category measured?
A free teardown runs your buyer prompts across every engine.
Get a free teardown