🔎 AI Member Screening

AI Member Screening evaluates new members using profile signals and, when needed, a model-assisted review. It complements Captcha and Spam Detection: captcha proves a human joined; screening asks whether that account looks risky before they settle in.

When screening runs

Two moments matter in the bot:

1. On join — A lightweight check can run for profiles with enough signal (long names, bios, linked channels, etc.).
2. First message — When a user matches your new user definition (for example “has never sent a message”), the bot can run a fuller pre-filter + model flow the first time they speak.

Configuration

1. Use /settings → 🤖 AI → 🔎 Member Screening, or send /ai_screening_settings as an admin
2. Toggle screening 🟢 Enabled or 🔴 Disabled

Actions

Flag for review

Suspicious cases generate an alert. Admins can use inline buttons on the alert to:

• Restrict (mute) the user for a duration
• Ban the user
• Dismiss the alert

Auto restrict

Automatically apply a mute when screening is confident there is an issue. Options include:

• Delete message — Optionally delete the triggering message when auto-restrict is enabled.

Notifications

Send screening alerts via:

• DM to admins, or
• A dedicated channel (you can bind a channel ID when prompted).

Troubleshooting

No alerts but you expect them?

• Confirm screening is 🟢 Enabled
• Check your New User Definition — screening keys off who counts as “new”
• Very minimal profiles may not trigger join-time screening until the first message

Too many false positives?

• Switch action mode to Flag for review instead of Auto restrict while tuning
• Use notifications in a private admin channel for faster context