Twitch banned words

What the Community Guidelines prohibit

The guidelines are organised by conduct, not vocabulary. The sections that govern what can be said in chat are headed “Civility and Respect”, “Hateful Conduct”, “Harassment”, “Sexual Harassment”, “Violence and Threats”, “Sexual Content” and “Spam, Scams, and Other Malicious Conduct”.

Hateful conduct covers slurs and attacks based on protected characteristics. Twitch lists these as race, ethnicity, color, caste, national origin, immigration status, religion, sex, gender, gender identity, sexual orientation, disability, serious medical condition and veteran status, with certain protections also applying to age. Hateful slurs are barred whether or not they are aimed at a specific person; denouncing a slur in context is allowed.

On a public list, Twitch is explicit: “We do not make our list of prohibited slurs public so as not to enable or facilitate evasion of this policy.” The one term the guidelines name outright is the N-word with a hard R, which “is never allowed under any circumstances”. Context is considered for other terms.

Ordinary profanity is not listed as a violation in the guidelines. What is a violation is how words are used: wishing harm on someone, implied threats, repeated personal attacks, unwanted sexual comments, and large volumes of repetitive unwanted messages (spam) are all named as examples.

Banned words in your own channel

Each channel sets its own words. Twitch’s moderation tools work in three layers:

• Blocked Terms and Phrases: words you add on the Moderation Settings page. A message containing a blocked term is not posted. A leading or trailing asterisk is a wildcard, so “shoot*” also catches “shooting”. A multi-word term blocks only messages containing all of its words, in any order. Each term can be Public (visible to your mods) or Private.
• AutoMod: holds messages it judges risky for a moderator to approve or deny before anyone else sees them. Twitch describes it as machine learning over natural language, not a word list you can read.
• Twitch’s own global filter: a “globally banned words filter”, on by default, containing words Twitch removes most often. Twitch does not publish its contents.

What streamers are expected to do

Twitch says streamers are expected to mitigate harassment and hateful conduct in their chat, and that it will not suspend streamers acting in good faith, using AutoMod, timeouts and bans, to remove abuse that third parties bring into the channel. Not using the tools at all, when hateful conduct appears on a channel, can lead to a suspension.

A practical corollary: keep your own chat logs. Twitch’s moderation view shows recent messages; a local archive lets you check what was said in a stream weeks later when a dispute or a sponsor question comes up.