Paul Bernabeu

I build measurement tools for AI search. One rule governs all of them: the numbers have to survive being shown to a client.

AI answers are not deterministic. Ask the same question twice and you get different brands, in a different order. Most tools in this space run a prompt once, report a single percentage, and call it a metric. Mine sample each prompt several times, report a confidence interval rather than a bare number, and refuse to call a month-over-month movement a trend unless it clears a significance test.

AI Visibility Tracker — GEO/AEO brand monitoring across ChatGPT, Google AI Overviews and Perplexity. Persistent history per project, so every run after the first tells you what actually changed and what was only noise.

• Method disclosed per engine. Every actor states which data is scraped, which comes from an official API, and who bills you for what.
• Cost estimated before the run starts, downstream charges included, and you are only billed for work that produced a result.
• Limitations documented up front, in the README, rather than discovered by you three runs later.

Based in France. Bug reports and feature requests go in the Issues tab and I read them.