# Twitch chat moderation

Bloopbot’s moderation filters, the commands your mods run from chat, and flows that react when a filter acts, with optional AI moderation.

## The filters

The filters are settings on the **Moderation** page, not flows, so they apply to every message. Each has its own thresholds, exemptions and response: warn, delete or time out.

The six filters are the blacklist, links, caps, symbols, emotes and repetition. Switch each one on when you want it, and watch chat for a while before turning on several at once.

The link filter and the blacklist see through disguises: a link spaced or spelled out (`example dot com`), written in look-alike or fancy letters, or broken up with invisible characters. Whatever a filter does, the message it caught is deleted; a first-offence warning only spares the viewer the timeout.

## Your mods, from chat

A moderator on Twitch is a moderator to Bloopbot straight away, with nothing to sign in to.

| Command | What it does |
| --- | --- |
| `!permit <username>` | Lets a viewer post a link past the link filter for a short time |
| `!filters enable <filter>` | Switches a filter on; `disable` switches it off |
| `!blacklist add <words>` | Adds a blacklist entry; `-regex` makes it a pattern |
| `!nuke` | Runs a bulk moderation job on chat, from the **Mod commands** pack; the **Nukes** page holds its settings |

The **Mod commands** pack also brings `!title` and `!game` for your stream information, and controls for the alert queue. `!permit`, `!filters` and `!blacklist` are built in and work without a flow. Anyone below moderator who types `!permit`, `!filters` or `!blacklist` gets no reply at all. Nukes may not be available on your channel yet.

## Flows for the rest

- **React when a filter acts.** The **Moderation acted** trigger starts a flow when a filter warns, deletes or times out, and can wait for a number of strikes: tell your mods in chat when a viewer reaches their third.
- **React to any ban or timeout.** The **Ban, timeout or unban** trigger fires whoever issued it, including from the Twitch dashboard.
- **Read what a message means.** An optional AI moderation step looks for harassment, hate speech, sexual content, scams and a rule you write in a sentence. Run it in watch mode for a stream first, and let it act once you trust it. It may not be available on your channel yet — see [set up AI moderation](https://bloopbot.com/docs/ai-moderation).

![The node library with the Groups topic selected: no triggers or conditions, and two actions — Enforce chat rules, your group used by 1 flow, leaving by Handled or Clean, and Screen reward text, your group used by 1 flow, leaving by Safe or Refunded.](https://bloopbot.com/docs/screenshots/node-library-groups.jpg)

_The node library’s Groups topic: Enforce chat rules is the group inside the AI moderation template, where you decide what each finding does._

## Check what happened

A flow’s **Run log** shows each path it took. The **Audit Log** shows who changed your channel’s settings, and the **Nukes** page lists bulk moderation jobs with their details. **Test fire** simulates a flow’s actions without timing out or banning a real viewer.

Editors you add on **Connections** can change the filters and flows from their own sign-in, and every change is recorded under their name — see [share the work with mods and editors](https://bloopbot.com/docs/editors). Editors may not be included on your plan.