Roblox Introduces AI-Powered Chat Rephrasing to Keep Conversations Civil

Image via Roblox

Roblox has rolled out a new AI-powered feature that automatically rephrases inappropriate language in real-time chat, replacing the old system of blocking text with hashmarks.

For a while now, the platform has relied on AI filters to catch language that violates its Community Standards. But anyone who has spent time in Roblox chat knows how frustrating it gets when half the messages look like a row of pound signs. The old approach got the job done on a surface level, but it made conversations clunky and hard to follow.

This new rephrasing system is Roblox's way of fixing that without loosening its safety standards.

Here is how it works.

Instead of showing "####" when a message gets flagged, the AI steps in and rewrites the offending text into something more appropriate. So a message like "Hurry TF up!" would now appear in chat as "Hurry up!" It is a small change, but it keeps the conversation moving naturally rather than grinding to a halt.

Transparency is built into the feature as well. Everyone in the chat sees a note when a message has been rephrased. The sender also gets to see exactly what was edited out.

And just to be clear, getting your message rephrased does not mean you are off the hook. According to Roblox, users who repeatedly break the profanity policy will still face consequences under the platform's existing guidelines.

Rajiv Bhatia, Vice President of User and Discovery Product at Roblox, explained: "Chat is central to how people connect, coordinate, and play on Roblox. Real-time rephrasing helps keep gameplay and conversations on track while guiding language toward what's appropriate. This approach reduces friction in chat while maintaining the standards that help keep our community civil."

The feature is currently live for in-experience chat between age-verified users in similar age groups.

It also supports all languages available through Roblox's automatic translation tools.

The company consulted members of its Teen Council during development, making sure the system actually reflects how teens talk in real life.

The groundwork for this was laid earlier. Back in 2025, Roblox introduced "proactive, real-time warnings" that nudged users to rethink their language before sending. As per dev, this experiments showed that in-experience chat notifications and time-outs led to a 5% reduction in filtered chat messages and a 6% drop in consequences from abuse reports.

Roblox is also upgrading its text filtering system. The improved filters are better at catching sneaky workarounds like leet-speak, where letters are swapped out for numbers or symbols to slip past automated systems.

These moves come against a backdrop of ongoing legal scrutiny, with lawsuits from LA County and Louisiana's Attorney General claiming the platform has “failed to do enough to protect children from predators.”