AI Moderation reviews chat messages with AI and takes automatic action when something breaks your rules. It goes beyond keeping a word list: it looks at intent, which makes it much harder to dodge the filter with spelling tricks.

Choosing a level

There are five steps from 0 to 4. Level 0 is unfiltered; 1–3 are preset strictness levels; 4 is personalised — you set a separate strictness for each topic: religion/politics, sexual content, hate speech, violence, bullying, drugs, weapons, spam, toxicity and more.

What actions can it take?

The default action is deleting the message.
A timed timeout can be applied, with a duration you choose.
A permanent ban can be enabled for serious violations.
A warning message can be posted to chat on a violation.
Usernames can be included in the filter if you want.

Word filter and spam protection

You can define your own rules in plain text, regex or obfuscation mode, each with its own action. Spam protection works on two layers: channel-wide (if the messages-per-second threshold is exceeded, switch to subscriber-only, enable slow mode, or clear chat) and per viewer (timeout or ban if one person exceeds the threshold). Both have a configurable cooldown and chat notification text.

FAQ

What if it takes a wrong action? Lower the level, or in personalised mode tighten only the topic that's causing trouble.
Does it work on Kick and YouTube? Yes, the platform selection is yours.
How is it different from Banned Words? That's a plain word list; this evaluates context and escalates actions.