YouTube's blocked words setting is a content filter. Creators sometimes search for it as YouTube banned words. In the interface the control is blocked words, and it is the right tool when a particular word or phrase should not go straight through.

It fails as a complete moderation policy because it only matches text. It does not read meaning, remember the account, or stop the next comment that uses different wording. A channel that relies on a list alone will keep meeting comments the list never named.

How to use blocked words today

In YouTube Studio on the web, open Settings, then Community, then Content controls. Add words or phrases separated by commas and save. YouTube's help documents the same Content controls in the YouTube Studio app under Community. Recheck the path on the day you use it.

Blocked words can apply to comments, live chat and Community posts. That breadth is useful. It is also the first limit: the control still acts on the matching text, not on the person who wrote it. A later comment from the same account is judged on its own words.

When a keyword list is enough

Use blocked words for tokens you never want in public: a slur you will not host, a spam phrase that keeps arriving, or a private detail you do not want repeated. A short, stable list of those is a sensible native control.

Do not ask the list to stand in for judgement. A word that is abuse in one sentence can be a quotation, a video title or an in-joke in another. A list cannot tell those apart. Lengthening the list until it matches everyday vocabulary will hide good comments and still miss the next invented spelling.

Why matching text is not the same as moderating a person

Three gaps show up quickly:

  • Meaning. The same letters can be hostility, reporting, or a joke. A list has no view of the thread.
  • Evasion. Spacing, homoglyphs and fresh phrasing walk around a stored word. YouTube's list and a local keyword rule both read the text they are given; neither promises to catch every disguise.
  • The next comment. Blocking a phrase does not change how the account's following comments are handled unless you take a separate account action.

Hiding a user from the channel is that separate action. It is not a keyword. It is also not temporary standing. YouTube has no dated probation for an account.

An illustration of the Rules screen. A keyword is one kind of rule, not the whole policy.

What The Litter Tray adds without pretending a list is enough

The tray can run a keyword or safe-pattern rule without spending an AI credit. A keyword is a literal whole word, not a fragment sitting inside a longer ordinary word. A new keyword must be at least four characters, because shorter needles match more than a creator usually meant. Homoglyph and spacing evasion stay out of scope; the editor is supposed to say so rather than imply the rule is cleverer than it is.

A word or pattern rule that holds is always set to Test, so it asks first. Only the built-in policy's six categories can be set to Auto.

When the job is judgement a blocked-word list cannot make, a plain-language AI rule can ask you in Needs review before it holds a comment. The creator still chooses the action. The classifier does not invent authority the rule was not given.

What this still does not do

  • It does not catch every inventive spelling or look-alike character.
  • It does not prove two accounts are the same person.
  • It does not put a viewer on probation when a word matches. Probation is a separate standing, set by the creator or by the setting that starts probation after comments held on Auto.
  • It does not write replies: it posts only the replies the creator writes.

See plain-language rules, AI is optional, and Test or Auto.