Tellcheck for GitHub flags likely AI-generated pull requests and issues on GitHub, right on the page. One click, comments scored, reasons shown. Built for maintainers who read contributions for a living and are tired of burning review time on generated ones.
Coming soon: Firefox add-on under AMO review Privacy policy
Codeberg, Godot, OpenJDK, Rust, and Ghostty all restricted AI-generated contributions this year, and enforcing a policy means someone has to read every submission twice. Tellcheck for GitHub does the first read: a scan button on every PR and issue, a badge on every comment, and a breakdown of exactly which signals fired so you can judge for yourself.
The classifier is calibrated to a 5% false-positive operating point and abstains on text too short to judge. Every result is labeled a signal, not a verdict. Flagging a real person's writing as AI is worse than missing one bot, and the tool is built around that. The corpus, the calibration, and what it gets wrong are all written up on the measurement page, with a command to reproduce the numbers.
Free, with a cap of 30 scored texts a day per install. No account, no signup. There is no paid tier yet; if one ships later, this page and the extension will say so before it changes anything for you.
Nothing leaves your browser until you click scan. Scanned text is scored and discarded, never stored. No accounts, no tracking. The whole policy fits on one page.
The same detector is also available as a metered HTTP API for scoring text anywhere, not just on GitHub: Tellcheck API.