Automod stops the obvious junk before a human sees it. The goal is high signal: catch spam and raids without punishing normal members.
Modules to enable first
Spam: repeated messages, fast posting, and duplicate content
Invites: block third-party Discord invites (allowlist your own partners)
Links: block or allowlist domains; catch known scam and phishing hosts
Banned words: tiered word lists with wildcards
Mentions: cap mass mentions and @everyone abuse
Caps: throttle shouting in busy channels
Severity tiers
Each module has a severity: delete only, delete + warn, mute, or escalate to a case. Start light — delete and warn — then tighten if abuse continues. Actions flow into the same case system as manual moderation.
Exemptions
Add role and channel exemptions so staff, boosters, or a #memes channel are not caught by strict rules. This is what keeps automod from feeling hostile to regulars.
Test it
Post from an alt or ask a staff member to trigger a rule on purpose in a test channel. Confirm the action, the log entry, and the DM copy before you rely on it.