Analysis of each chat message

How it works

Analysis happens in three steps:

1. A banned word filter immediately flags severe insults.
2. A repetition detector catches spam and flood.
3. An artificial intelligence reviews messages in batches, along with their context, to understand the real meaning of a sentence — a disguised insult or a real threat won't slip by just because it avoids using a banned word.

What is targeted

• Personal attacks/insults directed at the person (not their dino).
• Hate speech: racism, homophobobia, sexism, etc.
• Harassment: ganging up on a player, following them from channel to channel, real-life threats.

What's next?

When a message is problematic, it gets flagged to the moderation team along with its context. A human decides on any potential punishment: no penalty is automatically handed out for chat statements.

The special case of the Global channel

The Global channel is also monitored against metagaming: announcing in global chat where a player or a species is ("T-Rex at the south river"). In this case, the reputation penalty is automatic, because the offense is objective. If you think it's unjustified, open a ticket: it can be refunded.

The names you choose

The names you give to your skins and dinos stored in the park are also filtered (banned word list + automated analysis). An offensive name is simply rejected: pick another one.