Answer first
What to know
- Start with Discord's built-in commonly flagged words, spam, and mention controls before adding a large custom keyword list.
- Send alerts to a private moderator channel and use timeouts only for rules where the match is reliable enough to justify an automatic penalty.
- Test exact matches, wildcards, exemptions, usernames, links, and common false positives with an ordinary member account.
- Use Guardy for connected cases, reversals, escalation, anti-nuke controls, and raid cleanup when native AutoMod is only one part of the response.
Discord AutoMod can block undesirable messages before members see them, alert moderators, and apply a timeout for selected custom rules. The difficult part is not finding the switch. It is choosing a rule set that catches obvious abuse without punishing normal conversation.
Use native AutoMod as the first enforcement layer. Add a moderation bot when you need connected case history, reversals, escalation, anti-nuke protection, or coordinated raid response.
1. Open Discord’s AutoMod settings
AutoMod is available to Community servers. Go to Server Settings > AutoMod. Discord currently provides commonly flagged word categories, custom keyword rules, spam controls, and mention-spam controls. The available controls can change, so use the labels in the dashboard as the source of truth.
Create a private alerts channel before enabling responses. Only moderators and administrators should see the attempted message, author, channel, and matched rule.
2. Enable the reliable built-in protections first
Begin with rules that require little interpretation:
| Protection | Starting action | Why |
|---|---|---|
| Commonly flagged words | Block and alert | Uses Discord’s maintained categories |
| Mention spam | Block and alert | Objective count-based behavior |
| Obvious message spam | Block and alert | Limits repeated or high-volume disruption |
| Known malicious invite patterns | Block and alert | Narrow pattern with a clear policy |
Discord’s predefined categories cover insults and slurs, sexual content, and severe profanity. Review the category against the server’s audience before enabling it. A private adult community, family community, and professional support server will not share the same boundaries.
3. Add custom keywords in small groups
Do not paste an enormous generic blocklist and assume more entries mean better protection. Context, language, reclaimed words, usernames, and partial matches create false positives quickly.
Group custom rules by response:
- Always block and alert: clear scam domains, raid phrases observed in your server, and prohibited invite patterns.
- Block without alert: repetitive nuisance phrases that are not useful to investigate.
- Alert for human review: ambiguous terms where context matters.
Discord supports wildcard matching in keyword rules. A pattern such as *example* catches the text
inside a longer word, which can also match harmless words unexpectedly. Test both the intended match
and at least five normal words around it.
4. Choose responses based on confidence
Blocking a message prevents exposure. Sending an alert creates work for moderators. Timing out a user is a direct penalty. Those responses do not need to be enabled together for every rule.
Use automatic timeout only when:
- the rule has a very low false-positive rate;
- members already know the policy;
- moderators can reverse the action;
- the timeout duration is proportionate;
- exemptions have been tested.
For ambiguous toxicity or context-sensitive language, alerting a human is safer than turning one word match into an automatic punishment.
5. Configure exemptions narrowly
Some channels legitimately contain text that other channels should block. A security-report channel may contain suspicious links. Moderators may need to quote a prohibited phrase while documenting a case. Add exemptions for the smallest role or channel scope that solves the problem.
Avoid exempting the entire staff category if only one incident channel needs it. A compromised staff account with broad exemptions can bypass the same controls intended to protect the community.
6. Test like an ordinary member
Administrator accounts bypass or experience controls differently. Use an ordinary member test account and verify:
- a blocked message never appears publicly;
- the member sees an understandable reason;
- the alert reaches the private channel;
- exemptions work only in their intended scope;
- wildcard rules do not catch common innocent words;
- mention limits catch the planned threshold;
- moderators can find and reverse an action;
- the audit trail names the rule that fired.
Repeat the test on mobile if moderators commonly respond there.
7. Connect AutoMod to a complete moderation process
Native AutoMod answers whether one message matches one configured rule. A moderation operation also needs case history, consistent escalation, reversals, staff accountability, join protection, and a plan for coordinated incidents.
Guardy’s AutoMod module handles filters and automated responses. Its moderation module connects cases, escalation, reversals, join gates, anti-nuke controls, panic mode, and raid cleanup. That connection is the main reason to use a broader system alongside Discord’s native layer.
Review alerts after the first 24 hours and again after one week. Remove rules that generate noise without helping moderators. A smaller rule set that staff trusts is safer than a large rule set they learn to ignore. When the threat is coordinated rather than one-message abuse, use the Discord raid response checklist instead of trying to turn a keyword rule into an incident plan.
