It’s detection-only right now. No CAPTCHA, no forced blocking. I’m specifically looking for feedback on detection quality and false positives before automating enforcement.