Skip to main content

​
Gaming and social chat moderation

Short answer: Tuteliq moderates gaming and social chat by reading the conversation, not a word list. It keeps competitive trash-talk intact while catching targeted harassment, coded slang and emoji, grooming of minors, and coordinated pile-ons, and it runs in real time so you can act on the verdict as messages are sent. That context is what a keyword filter and a single-message classifier cannot do.

​
Why gaming chat is hard to moderate

Game and social communities run on banter, in-jokes, and fast-moving slang. That breaks the two common approaches at once:
  • Keyword filters flag “I’m going to destroy you” in a lobby and bury your team in false positives, while missing the coded slang, emoji, and algospeak that real harassment actually uses.
  • Single-message classifiers score one line in isolation, so they cannot tell reciprocal banter from a sustained attack, and they miss harm that only makes sense across several messages.
Tuteliq scores the interaction: who is targeting whom, whether it is reciprocal, whether it is escalating, and whether a group is coordinating.

​
What it catches in game and social chat

  • Harassment vs banter. Two players swapping “get rekt” is banter. One account repeatedly targeting a player who is asking them to stop is harassment. Tuteliq weighs reciprocity, target, intent, and persistence. See Bullying and toxicity detection.
  • Coded slang, emoji, and evasion. Abusive users deliberately dodge filters with leetspeak, emoji payloads, homoglyphs, and algospeak. On an evasion benchmark, Tuteliq detected roughly 1.7x more coded-language cases than general-purpose moderation APIs. See Benchmarks.
  • Grooming of minors. Game chat and DMs are a common grooming surface. Tuteliq detects the trajectory (trust-building, isolation to another app, secrecy, meeting requests) across many turns, and reasons about age and role. See Grooming detection.
  • Coordinated abuse and pile-ons. A single insult is low severity; a group turning on one player, mass-reporting, or creating a side-channel to mock them is a different and higher-severity pattern.
  • Threats and self-harm-adjacent language, escalated appropriately rather than treated as ordinary game talk.

​
Built for real-time chat

  • Fast mode. Screen every message with fast mode (verdict_only: true) for low latency, then re-run only the messages that flag through standard mode for the full per-message detail your moderators need.
  • Multi-turn awareness without storing chat. Pass a continuation token from each call into the next to preserve trajectory awareness across a session. Tuteliq stores no user content.
  • Webhooks and routing. Attach your own external_id and customer_id so alerts route straight back to the right player, match, or tenant.
  • 27 languages with code-switching, so mixed-language lobbies are covered. See Languages.

​
Using it

For grooming and predatory-behaviour detection in game DMs, use /api/v1/safety/grooming with the conversation history.

​
Why platforms choose Tuteliq

  • Context over keywords, so healthy competitive chat stays intact and real harm still fires.
  • Coded-language resilience, roughly 1.7x the recall of general-purpose moderation APIs on evasion cases.
  • One API for many harms: toxicity, harassment, grooming, threats, and more.
  • Explainable: every result returns a rationale, categories, and a recommended action for moderator review.
  • Privacy-first: no user content is stored.

​
Frequently asked questions

Do you have a content moderation API for gaming chat? Yes. Tuteliq moderates game and social chat in real time, reading conversation context to separate trash-talk from harassment and to catch coded slang, grooming, and coordinated abuse. Can it handle gaming slang, emoji, and filter evasion? Yes. It is maintained against an evolving lexicon of coded terms, emoji, leetspeak, and algospeak, and detected roughly 1.7x more evasion cases than general-purpose moderation APIs on a benchmark. Is it fast enough for live chat? Yes. Use fast mode (verdict_only: true) for low-latency screening of every message, and reserve full per-message analysis for the messages that flag. Can it detect grooming in game DMs? Yes. The grooming detector tracks predatory trajectories across many turns and reasons about the age and role of each participant. What languages are supported? 27, including all 24 EU official languages plus Ukrainian, Norwegian, and Turkish, with code-switching support.