Your AI visibility score moved 20 points. Nothing changed.

Ask ChatGPT the same question ten times and you get a different set of sources most of the time. Same-day source overlap between two runs of one prompt is 23 to 51 percent depending on the engine. Trackers show you one run and call it a score. Nobody tells you which page change actually got you cited, or whether the answer is “we can’t tell yet”.

Keoph runs randomized experiments on groups of your docs and pages, measures the result on Google’s and Bing’s own per-page AI reports plus repeated sampling, and reports an effect with an interval, including “inconclusive”. Built for developer-tool companies.