AI Moderation

AI Moderation uses your configured AI provider (OpenAI, Claude, Gemini, DeepSeek, or Grok) to read each chat message and automatically flag content that violates the report-reason categories you choose. It can censor offending messages automatically or queue them for a moderator.

Navigate to Moderation > AI Moderation

How It Works

Each message is sent to your moderation AI with a strict prompt listing only the categories you enabled.
The AI decides whether the message clearly violates one of those categories, and if so, which one.
Your selected Action is then applied — the message is censored and the report auto-resolved, or it is flagged for manual review.
The categories are your report reasons — the same reasons members pick when they report a message.
If the AI provider is unreachable, the message is left as-is (fails open).

Step 1 — Set Up an AI Provider (Required)

AI Moderation does not work on its own — it relies on an AI provider configured in your AI settings.

Go to General Settings > AI and enable at least one provider by entering its API key. The AI Settings article lists exactly where to generate a key for each provider (OpenAI, Anthropic/Claude, Google Gemini, DeepSeek, Grok).
In the same page's Task Configuration section, set the Content Moderation task to Default or to a specific provider. If this task is set to Disabled, AI Moderation will not run.

Tip: You can route moderation to a cheaper, faster model than the one powering your chatbot — assign each task its own provider in AI Settings.

Step 2 — Configure AI Moderation

Go to Moderation > AI Moderation:

Setting	Description
Enable AI Moderation	Set to Yes to turn it on.
Categories to Moderate	Tick the report-reason categories the AI should enforce (for example Spam, Harassment, Hate Speech). The AI only flags messages that clearly match one of these.
Action for Flagged Content	Censor & Auto-Resolve replaces the message and closes the report automatically. Flag for Review keeps the message visible and creates a report for a moderator.

Click Update to save.

Note: If no categories are selected, AI Moderation stays inactive. The category names come from your report reasons and can be renamed or translated under Appearance > Language.

Reviewing Flagged Content

When the action is Flag for Review, flagged messages appear under Moderation > Flaged Content together with the reason the AI matched. See IP Access & Flagged Content.

Cost & Performance

Every moderated message is one AI request, billed by your provider per token. Short chat messages are cheap individually, but high-traffic chats add up — monitor usage in your provider dashboard.
Choose a small, fast model for moderation to keep latency and cost low.

AI Moderation vs Image Moderation

AI Moderation handles text messages. To scan uploaded images, use Image Moderation. The two can run at the same time.

AI Settings — provider setup and where to get API keys.
Image Moderation
Moderation Tools
IP Access & Flagged Content