What it does
Only loaded text replies are examined, and the estimate is not proof of authorship.
Browser & computer use
Mark or collapse X replies that resemble generic AI-written paraphrases of the original post.
Only you can see your notes.
Screenshot unavailable. Open the experiment ↗
Only loaded text replies are examined, and the estimate is not proof of authorship.
First, decide whether you want to highlight suspected bot comments or hide them completely. Have a developer build the extension files and load them into Chrome. You will paste an access key from Vercel into the extension menu so it can send post text to the AI model.
On X, click the crossed-out bot icon at the top of a post to begin. The tool checks visible replies as you scroll down. Remember that these scores are only estimates, so real human comments might still be flagged by mistake.