Set up Discord AutoMod: block bad words, spam, links and phishing

Once your server grows, the same problems keep coming back: insults, ads for other servers, message floods and fake Nitro links. A moderation team cannot read along around the clock. A well configured Discord AutoMod catches the worst of it automatically, before anyone sees it or clicks on it.

In this guide you learn what Discord's own AutoMod can do, where a bot AutoMod does more, and how to combine both. At the end you have a word filter, rules for links and invites, spam protection and a phishing filter that fit your server.

Discord's own AutoMod vs. a bot AutoMod

Discord has a built-in AutoMod. You find it in your server settings under AutoMod. It offers rules against mention spam, suspected spam, commonly flagged words from Discord's own lists, and custom keyword rules. As a response, Discord can block the message, send an alert to a channel or time the member out.

The big advantage: Discord blocks the message before it shows up in the channel. The downside: the rules are rigid. There are no warnings with a case file, no escalation, no checks across several messages and no lists of known scam sites.

A bot AutoMod like Lizbot's reads every message, checks it against your rules and acts afterwards. That lets it do more: it spots spam across several messages, checks links against scam lists, sees through spelling tricks and saves every punishment as a case your team can look up later.

You don't have to choose. Lizbot can also create your word filter and mass mention protection as native Discord rules. Discord then blocks the message right away, and Lizbot handles everything Discord can't.

Who is exempt from AutoMod

AutoMod never acts on mods and admins. You decide which roles count as such in the dashboard under Settings with Admin roles and Mod roles. There you can also add Trusted users (IDs), for example a partner who posts links regularly.

On top of that, every rule has its own Exempt roles and Exempt channels fields. That way you can allow links only in #media without turning the filter off everywhere.

1. Load an AutoMod level

Open AutoMod in the dashboard and stay on the Overview tab. With Load AutoMod level you apply a ready-made baseline:

• Relaxed: only invite spam and mass mentions, for small, trusted servers
• Standard: word packs against insults, hate speech, threats and scams, plus invites, spam, mentions and raid protection with kick
• Strict: adds sexual content, drug dealing, a link whitelist and ban on raids
• Family / kids: mild swearing too, links only from a small whitelist, no invites
• Business: factual, blocks spam and scams, allows links from work tools

For most communities, Standard is the right start. Then pick an AutoMod log channel that only your team can see. Every match lands there with the message, rule and action. How to set up logs for your whole server is covered in a separate guide.

Below that is the switch Sync with Discord's native AutoMod. Turn it on if messages should never show up in the first place. The option Also use Discord's own word lists blocks everything on Discord's lists for profanity, slurs and sexual content. That also catches ordinary swearing, so only use it on servers that really want that.

2. Set up the word filter

On the Word filter tab you work with word packs maintained by Lizbot: Insults, Mild swearing, Hate speech, Sexual content, Self-harm encouragement, Threats, Drug dealing, Scam phrases, and Nazi slogans and codes. Every pack has severity levels from low to high and, if you want, its own action. So you can punish hate speech with a timeout while a mild insult is only deleted.

Under Languages you choose which languages are checked, for example English, German and Turkish. Under Your own words you add terms that are only a problem on your server. An asterisk at the start or end works as a wildcard, several words form a phrase.

The filter catches common tricks automatically: leetspeak like 1d10t, repeated letters, special characters, and dots or spaces between letters. To keep harmless words from matching, there is an Allow list, for example for place names or technical terms.

Use the Test field before saving. You type a sentence and instantly see whether and why it would match.

3. Control invites and links

On the Rules tab you find Invite links. Enable the rule and switch on Allow invites to this server, so members can still share invites to your own server.

For Links you pick a mode:

• Blacklist: block listed domains only
• Whitelist: allow listed domains only
• Block all links

For open communities, a blacklist together with phishing protection is usually enough. A whitelist makes sense for kids or business servers. If a GIF provider like tenor.com is on the whitelist, Lizbot automatically allows the other providers from Discord's GIF picker too, so GIFs don't suddenly disappear.

4. Stop spam and mass mentions

Under Mass mentions you set how many mentions a single message may contain. A value around 5 to 6 suits most servers.

On the Spam & raid tab you configure the spam filter: how many Messages are allowed within how many seconds, and from how many Identical messages Lizbot steps in. The Standard level uses 6 messages in 5 seconds and 3 identical messages. With Also delete the earlier spam messages, Lizbot cleans up the rest of that member's messages from the time window on a match. The Rules tab also has the Capital letters filter.

5. Turn on phishing protection

Fake Nitro gifts and Steam pages are the most common scam on Discord. On the Phishing tab, enable Use public scam lists. Lizbot reloads these lists every six hours, so there is nothing for you to maintain.

Lizbot also detects lookalike addresses that are easy to confuse with real Discord or Steam domains. Under Custom blocked domains you add sites that keep showing up on your server. Phishing is checked before all other rules.

6. Choose actions

Every rule has an Actions field. You can pick several: Delete, Warn, Timeout, Kick and Ban. Every punishment is saved as a case that your moderation team can review with warnings and case files later. For timeouts you set the Timeout duration. With Log matches, the match is posted to the AutoMod log channel.

Start mild: delete plus warn is enough for most violations. Timeouts make sense for spam and mass mentions, bans only for clear cases like scams. Which punishment fits when is explained in Timeout, kick or ban. So nobody escalates from warning to ban within seconds, Lizbot hands out at most one punishment per member every 15 seconds. Deleting and logging still continue.

7. Add AI review and raid protection

The Studio plan adds the AI tab. AI review catches insults, harassment, scams and more, even when nothing in the word filter matches. Messages are sent to the AI for this, which you should mention in your server rules.

From Pro onwards there is also the Scam shield tab. The AI scam shield only checks messages that a pre-filter flags as suspicious, for example ones with unknown links or free Nitro bait. For the first 7 days it only observes, and your team corrects every detection with Correctly detected or False alarm.

Against join waves and alt accounts of banned users, use the Spam & raid and Alt accounts tabs. How to set them up is covered in the guide Set up raid protection.

Recommendations by server type

Checklist

• AutoMod is enabled and a level is loaded
• an AutoMod log channel for the team only is set
• mod and admin roles are set in Settings
• word packs and languages fit your server
• invite links are blocked, your own invites allowed
• phishing protection with public scam lists is on
• the word filter has been tried in the test field

Common mistakes

• Starting too strict: Setting ban as the action on day one loses members to false positives. Start with delete and warn.
• No log channel: Without a log, your team doesn't know what AutoMod does and can't fix false positives.
• Missing mod roles: If the roles aren't set in Settings, Lizbot treats your team like regular members.
• Turning on Discord's word lists blindly: They also catch everyday swearing and can't be fine-tuned per word.
• Blocking all links: In communities that share clips and videos, that causes frustration. Use a blacklist plus phishing protection instead.