Strict automoderation without false positives

Automoderation that is too soft lets spam through; too harsh, it times out a member sharing three links in a row to help someone. Good news: with a few settings, you get both, strict with spammers and tolerant with humans.

Start with realistic thresholds

For “Message burst”, 6 messages in 10 seconds is a good start: an enthusiastic member almost never reaches it, a spam bot gets there in one second. Raise the threshold in fast-chat channels rather than lowering it everywhere.

Exclusions, your best tool

Every rule has its Groups (who is affected) and Exclusions (ignored roles and channels). Exclude staff, bot channels, and link-sharing channels if you have them. An invite rule can allow invites from partner servers.

Enlarge

Editing an automoderation ruleOriginal image

Heat mode: sanction repetition

Under “More options”, heat mode accumulates points on each detection instead of sanctioning right away. Example: 5 points per detection, threshold at 15, 30-second decay. Someone typing fast once isn’t sanctioned; someone who keeps going reaches the threshold.

Alert instead of punish

Some words need human judgement: “threat”, “harassment”, the name of a frequently targeted member. For those, use the Alerts tab: Spider notifies staff in a channel, without sanctioning. Keep lists short and a delay of at least one minute between alerts.

Enlarge

Automoderation alerts and destination channelOriginal image

Golden rule: automate whatever deserves the same answer every time (ads, spam). Alert for everything that depends on context.