Knowledge Base Article

AI Moderation: Moderator workflows

Access all tools from Community HomeAdmin Moderation. The dashboard has eight tabs: Drafts (review before publication), Spam (confirm or clear AI-rejected content), Abuse (user-reported content), Moderation (primary review queue — approve, reject, view AI analysis), Appeals (approve or reject appeals), Archives (restore if needed), Tags, and Guides.

Moderation Queue

Manage Content > Moderation

Shows pending content with columns: Title, Author, Date, Status, and AI Verdict (e.g. "Acceptable (60%)", "Spam (90%)"). Click the ⓘ icon on any verdict to open the AI Moderation Analysis modal.

Filters: Moderation Status (Unmoderated, Rejected) | Published Status | Found In (Forums, Ideas, Blog Comments, KB Comments, Event Comments) | Reject Reason (Spam, Abuse, Other) | Contains Media (Images, Videos, Attachments) | Keyword search, Author filter

Use filters to batch-review similar content, e.g. "Contains Media" + "Images" to review all flagged images together.

AI Moderation Analysis Modal

Shows the verdict (Approved/Rejected) with confidence percentage, AI provider, and timestamp; the Reason Code(s) applied with technical ID and user-facing message; and the AI's plain-language reasoning. Low confidence (below your threshold) means uncertainty, reasoning is provided. High confidence (above 85%) indicates strong evidence.

Appeals Queue

Manage Content > Appeals

Keep the default "Pending" filter active to focus on decisions awaiting your action.

Appeal Detail Modal: shows the user's appeal category and reason text, the original post content, the AI analysis (verdict, confidence, reason codes, reasoning), and Reject Appeal/Approve Appeal buttons.

Consider approving if AI confidence is low, and the user provided valid context, the content fits community norms, or it's a clear false positive (jargon, technical language).

Consider rejecting when AI confidence is high (above 85%) with clear reasoning and the user's context doesn't address the violation, or promotional or spam intent is evident.

Updated 1 month ago
Version 2.0
No CommentsBe the first to comment