A properly configured auto-mod handles 85% of moderation issues without human intervention. Here's the exact setup that works — from spam detection to raid lockdown.
Layer 1: Spam Detection
Spam is the most common moderation issue and the easiest to automate. Set thresholds, not keywords. Keyword filters are gamed in minutes. Rate-based detection is far more effective.
| Rule | Threshold | Action |
|---|---|---|
| Message rate | 5 msgs in 5 seconds | Timeout 10 min |
| Duplicate messages | 3 identical in 5 min | Delete + warn |
| Mention spam | 5+ mentions in 1 message | Delete + timeout 1h |
| Link spam | 3+ links in 2 min | Delete + warn |
| Caps abuse | 80%+ caps, 20+ chars | Delete |
Layer 2: Raid Protection
Raids are coordinated mass-joins designed to spam or overwhelm a server. The key is detecting the join velocity spike before the spam starts — not after.
- →Set a join rate alert: 10+ joins in 30 seconds = auto-lockdown mode
- →Lockdown mode: pause verification, restrict new members to read-only for 10 minutes
- →Auto-ban accounts under 7 days old during an active raid (toggle off after)
- →Enable CAPTCHA verification only during lockdown — not permanently, it kills conversion
server.ninja's raid protection activates in under 1 second. It detects join velocity spikes before the first spam message hits. You'll see the alert before your members see anything.
Layer 3: Word Filters
Word filters should be surgical, not comprehensive. Trying to filter every slur creates thousands of false positives that alienate good members. Focus on three categories only.
- →Category 1: Obvious slurs — zero tolerance, immediate ban
- →Category 2: Context-dependent terms — warn + human review, not auto-ban
- →Category 3: Server-specific banned topics — delete only, no punishment unless repeated
- →Never filter common words with double meanings — the false positive rate destroys trust
Layer 4: New Member Restrictions
New members are responsible for 80% of spam and raid activity. A brief restriction window eliminates most of it without affecting legitimate members.
| Restriction | Duration | What it prevents |
|---|---|---|
| No external links | 24h | Phishing and spam links |
| No embeds | 12h | NSFW image spam |
| No mass mentions | 48h | Mention spam raids |
| Slow mode (3s) | First 7 days | Drive-by spam accounts |