All docs
Moderation & Safety

Auto-Moderation

Let Hez delete spam and rule-breaking messages by itself

Auto-moderation reads every message as it's posted and reacts before a human has to. It isn't one filter — it's a set of independent detection rules you turn on one at a time, each with its own thresholds and its own response.

Every rule can escalate: a first hit might just delete the message, a fifth hit from the same person in a short window might time them out. You decide where that line sits per rule.

Turning it on or off

Every server chooses for itself — switch Auto-Moderation on or off from your dashboard, under Moderation & Safety. When it's off: messages will no longer be checked automatically.

Setting it up

  1. 1

    Turn on one or two rules to start

    Scam links and mass mentions are the safest starting point — they rarely false-positive on real conversation.

  2. 2

    Add your staff to the exempt roles

    Do this before you turn anything on, not after someone gets timed out for testing.

  3. 3

    Watch the automod hit log for a few days

    Confirm it's only catching what you expect before adding stricter rules like caps or wall-of-text.

  4. 4

    Set an escalation ladder for repeat offenders

    Delete on the first hit, warn on the third, timeout after that — tune it to how forgiving you want to be.

Escalation and response

Every rule's response is configurable independently: delete the message, warn the author, time them out, or escalate further after repeat hits within a window you set. You aren't limited to "delete and done" — a rule can ramp up automatically for someone who keeps tripping it.

Raid Mode (part of the Protection module) can temporarily tighten every automod rule's strictness at once during an active raid, then relax back to normal when it ends — you don't have to hand-tune rules mid-incident.

Exemptions

Add your moderators — and any bots that legitimately post links or repeat content, like a notifications bot — to the exempt roles list. Without it, your own team's testing and normal bot activity gets caught by the same rules meant for spammers.

Detection rules available

Start with a couple and watch the automod hit log for a few days before adding more — a rule that's too strict deletes real conversation, not just spam.

Mass spam

The same message (or near-identical ones) posted repeatedly, including across multiple channels at once.

Scam / phishing links

URLs matched against a maintained list of known scam and phishing domains — Discord nitro scams, fake crypto giveaways, and similar.

Mass mentions

Pinging an unusual number of members or roles in one message, a common raid tactic.

Repeated spam

One member sending the same content over and over in a short window, even if it isn't cross-channel.

Invite spam

Unsolicited Discord invite links to other servers.

Caps spam

Messages that are excessively upper-case.

Emoji spam

An excessive number of emoji packed into one message.

Link spam

A general flood of links, distinct from the scam-specific rule above.

Banned words

Your own list of disallowed words or phrases.

Zalgo / unicode abuse

Messages using combining-character tricks to distort or overflow the display.

Attachment filter

Block specific file types, or attachments over a size you set.

Sticker spam

Excessive stickers, mirroring the emoji-spam rule.

Wall of text

Unusually long single messages.

Spoiler spam

Excessive use of the spoiler-tag formatting.

Channel burst

A raid-style burst of messages across several channels at once, rather than repetition in one channel.

Nickname filter

Flags nicknames matching a pattern you define — slurs, impersonation attempts, and similar.

Anti-hoist

Catches names that use leading symbols purely to sit at the top of the member list.

Commands

No dedicated slash commands — everything for Auto-Moderation is set up on the dashboard.

Tips & things people get wrong

  • It's easy to be too strict on day one. A quiet rollout beats an angry server full of "why was my message deleted" pings.
  • Caps-spam and wall-of-text rules are the two most likely to catch genuine, if enthusiastic, members — turn those on last.
  • Auto-moderation and Protection's Raid Mode work together: Raid Mode's strictness multiplier only has automod rules to tighten if you've actually turned some on.

Ready to try Auto-Moderation?