Roblox and Discord Chat Safety, a Parent's Guide

Published: September 15, 2026. Updated: October 2026 by Andrew Apell, who builds PureWord and previously built Flookup Data Wrangler

Quick Answer

Short answer: Game chat filters miss disguised profanity. Open Roblox or Discord in Chrome on desktop with the free PureWord extension switched on. It folds misspellings and symbols back to base words before matching, and custom word pairs close the gaps that remain. Voice chat and mobile apps stay outside its reach.

Game chat is where children meet language no parent approved. Roblox hosts millions of young players. Discord servers mix ages freely, and both move too quickly for moderation to catch everything. The deeper problem is that the filters inside these platforms were built for a simpler threat than the one children actually meet.

This guide explains how evasion works, shows a gallery of bypass patterns with genuine outcomes from the PureWord engine, recommends word lists by age group and states the limits plainly. I wrote it for parents who want understanding first and installation second.

How Evasion Defeats Built-in Filters

Built-in chat filters compare each message against a list of banned spellings. Players learned long ago that the list is literal, so evasion is now routine: swap a letter for a symbol, stretch a vowel, separate letters with spaces or borrow a lookalike character from another alphabet. Each trick produces text that reads clearly to a child and means nothing to a keyword list.

A text filter that normalises before it matches behaves differently. Inflections stem to base words, digits standing in for a letter fold back, and hyphens, full stops and underscores come out. Measured across every printable character, that normalisation is narrower than the marketing usually claims, which is why the gallery below lists the misses as carefully as the catches.

Method: I ran the bypass patterns below through the current PureWord engine with the standard English word list in Chrome on a desktop, in the default masking configuration, and re-measured on October 3, 2026 with a benchmark anyone can re-run. Every pattern is listed, including the ones that were not caught.

Bypass pattern Example shape Outcome
Symbol for letter a listed spelling such as sh!t, written out in words here Filtered, but only because that exact spelling is in the word list. Symbols do not fold in general.
Number for letter d4mn with a digit for the a Filtered. Digits fold back, and the fold covers 4 and @ for the letter a.
Doubled symbols ass with a repeated symbol Filtered where the exact spelling is listed.
Asterisk split d#mn with a hash for the a Not filtered. Of all 94 printable ASCII characters, only 4 and @ substitute for a letter at all.
Inflected insult damns with a plural ending Filtered. Endings stem to the base word at match time.
Lookalike alphabet Cyrillic o and Greek omicron standing in for Latin o Not filtered for most letters. Of 39 Unicode lookalikes tested, 4 fold away and all 4 stand in for a.
Spaced letters d a m n with spaces between letters Not filtered automatically. Add the exact spaced spelling as a custom pair.
Stretched vowels daaaamn with a repeated vowel Not filtered automatically. Add the exact stretched spelling as a custom pair.
Direction overrides reversed text with Unicode controls Not filtered. Measured, the override is not removed before matching.

Four of the nine patterns fall to normalisation alone and five are missed, a higher miss rate than the pattern of most published comparisons suggests. The misses fail for two different reasons: the letters never assemble into a recognisable token, or the disguise uses a character the engine does not fold. Because custom word pairs match literally, they cover both, so the combination of automatic folding and a short custom list is what actually closes the gap.

Word Lists by Age Group

The right list depends on the child. A young child repeats what appears on screen, so breadth matters more than precision. A teenager needs the standard list plus the slang of their servers, and overblocking mild language can push them to disable the filter entirely, so restraint matters there.

Young Children, Roughly Ages Six to Ten

Start with the full standard list and add playground language, stupid, ugly, shut up, dumb and hate you. These are not profanity in the strict sense, but young children echo them at school and at home. Add character names or insults from the specific games your child plays, since game specific taunts bother children more than generic swearing.

Teenagers, Roughly Ages Eleven and Up

Keep the standard list and add server slang rather than mild words. Ask which terms bother them and add those together, because a teenager who helped choose the list will keep the filter switched on. Skip playground words at this age, since masking every mild term trains teenagers to bypass the tool instead of gaining from it.

Setup on the Web Clients

Add PureWord from the Chrome Web Store. It is free, installs in a single click and asks for no account. Open roblox.com or discord.com in Chrome on desktop and confirm the popup toggle shows filtering as on for the page. Chat messages are filtered as they arrive, including history that loads while scrolling.

Enter custom pairs in the options page as word, replacement per line. Enable Parental Lock so settings stay password protected and masked words cannot be clicked to reveal the original on a shared computer.

What Stays Outside the Filter

Voice chat carries no page text, so nothing there is filtered. Mobile apps run outside Chrome, so the Roblox and Discord apps on phones and tablets need their own platform controls. Direct messages on the web clients are filtered like any other chat text, which surprises some parents pleasantly, but voice channels in the same servers are not.

FAQ

Why do swear words get through Roblox and Discord filters?

Built-in filters match dictionary spellings. Players disguise words with symbols, swapped letters and spacing, and simple keyword lists miss them.

What words should I block for a young child?

Start broad: standard profanity plus playground insults such as stupid, ugly and shut up. Young children repeat what they read, so cover mild language too.

Does this cover voice chat and mobile apps?

No. A text filter sees page text only. Voice chat is untouched and mobile apps are outside Chrome, so use the web clients for filtered chat.

Is my data safe?

Everything runs on your device. PureWord does not collect data, does not track you and does not send your browsing content anywhere. See the privacy policy for details.

Summary

Game chat evasion beats keyword lists by disguising words faster than lists update. Normalisation before matching reverses most disguises, custom pairs cover spaced and stretched spellings, and age tuned lists keep the filter appropriate without provoking bypass. Use the web clients in Chrome, lock the settings and treat voice chat as a separate conversation with your child.

Filter Game Chat as It Arrives

PureWord normalises evasions before matching on the Roblox and Discord web clients. Free and private.

Add to Chrome
Project Slidecut