01Why the usual way fails
Keyword filters only catch the exact words you list, so people get around them with numbers instead of letters, odd Unicode characters or polite-sounding insults. A message can look harmless on its own and still be part of a longer harassment pattern that only makes sense with the history. Scam posts like fake giveaways and phishing links spread fast, and members click before anyone removes them. Communities that speak many languages make manual review even harder, because no moderator team covers them all.
02How Supervisor helps
Supervisor is a moderation service that reads the intent of a message, not just its keywords, and runs around the clock. It flags spam, harassment, hate speech, NSFW content and scams in real time, and it can check the conversation history when a message only makes sense in context. It handles more than 100 languages, sees through l33t-speak and Unicode tricks, and also scans images and video clips. You can add it to Discord as a bot or call it through its API from your own app.
“Moderation that reads intent, not just keywords.”
“Supervisor works around the clock to keep your community safe.”
03Step by step
- Sign up and try the demo box to see how a message gets classified.
- Add the Supervisor bot to your Discord server using the Discord integration guide.
- Choose which categories matter to your community, such as spam, scams or harassment.
- Watch the dashboard logs for the first few days to see what gets flagged.
- Adjust your settings, and keep human moderators for appeals and borderline cases.
04Tips
- Pay special attention to scam and phishing links, since they do the most damage in the shortest time. Fake gift and giveaway posts are flagged as they are posted, before members can click them.
- Do not rely only on bad-word lists. Many insults use no swear words at all, and implicit detection is built to catch sarcasm and backhanded attacks.
- Write down clear rules and pin them before you turn on automated moderation. When members know the rules, flagged messages are easier to explain and fewer people feel treated unfairly.
- Keep a human in the loop. An automated tool can remove the obvious problems at night, but your moderators should still review edge cases and talk to members when a situation needs judgment.
05Questions
Which languages does it support?
More than 100 languages, and Unicode is supported.
Source: supervisor.ggCan it catch harassment that builds over several messages?
Yes. It reads the conversation history and flags intent that single-message checks would miss.
Source: supervisor.ggWhat content categories does it check?
Categories include profanity, toxicity, harassment, hate, insults, sexual content, violence, self-harm, spam, scams and illegal activity.
Source: supervisor.gg