The smart check asks whether the message harasses, bullies, or personally attacks someone. Friendly banter in an ongoing conversation can score lower when recent context shows the tone is mutual.
Typical catches
- Direct insults and pile-ons against one member.
- Repeated demeaning comments after someone asked to stop.
- Personal attacks disguised as jokes when context shows they are not mutual.
Outcomes
Above your action threshold: message removed and strike points. Between review and action thresholds: case card for mods, message stays up until someone acts.
Tuning
Turn the category off in /settings if you handle harassment manually, or raise thresholds if your community's banter runs hot. Child safety and self-harm checks still run separately.