Twitch's Suspicious User Detection uses machine learning to analyze account signals and chat behavior to identify users who create new accounts to circumvent channel-level bans, flagging them as "likely" or "possible" evaders for moderator review.

Launched in late November/December 2021 as a direct response to the 2021 hate raid crisis (#TwitchDoBetter, #ADayOffTwitch protests). The system compares account characteristics and chat behavior against banned accounts from a given channel. Users flagged as "Likely" evaders have messages automatically hidden from general chat; "Possible" evaders are flagged but can still chat. The tool was enabled by default on all channels. Twitch explicitly warned that "no machine learning will ever be 100% accurate" and designed the system so it does not automatically ban anyone — all final decisions remain with human moderators.