Skip to main content

👿 Spam Detection

Automatically detect spam sent by new users. This feature uses Fren One's custom AI model to identify spammy or scam-like messages and either removes them instantly or applies configured moderation actions, ensuring your chat remains free from unwanted distractions.

Overview​

Fren One's Spam Detection leverages a custom AI model deployed by Fren One specifically trained to identify spam patterns in Telegram communities. This is not one of the external models that require API keys from users, which means:

✨ Zero cost to you - No inference tokens consumed from your OpenAI or ASI:Cloud accounts
✨ Always available - Works independently of your API key configurations
✨ Optimized for Telegram - Trained specifically on Telegram spam patterns
✨ Real-time protection - Continuous monitoring with instant response

No API Keys Required

Unlike Smart Triggers or other AI features that use OpenAI/ASI:Cloud APIs, Spam Detection runs on Fren One's own infrastructure. You don't need to provide any API keys, and it won't consume your inference credits.

How It Works​

Fren One's AI continuously monitors all messages from new users in real-time. It analyzes the content for typical spam indicators like:

  • External group mentions: Tagging multiple users to get attention
  • Excessive links or emojis: Overuse of promotional links or emoji spam
  • Scam keywords and context: Terms like "airdrops," "giveaways," "send me your wallet," "guaranteed profit"
  • Repetitive messages: Same or similar content posted multiple times
  • Suspicious forwarding patterns: Mass-forwarded messages from other groups
  • Unusual message structures: Patterns often associated with bots and automated spam

The AI model has been trained on thousands of spam examples and continuously improves its detection accuracy.

New Users Only

Spam Detection specifically targets new users to protect against the most common source of spam while minimizing false positives from established community members. Define "new user" criteria in your Filter settings.

Configuration​

To access Spam Detection settings:

  1. Use the /settings command in your chat
  2. Navigate to 🤖 AI → 👿 Spam Detection
  3. Use the toggle to turn the feature 🟢 ON or 🔴 OFF

Spam Detection Settings​

Enabled/Disabled Toggle​

Toggle AI-powered spam detection for this community.

  • 🟢 Enabled: Spam detection is active for new users
  • 🔴 Disabled: Spam detection is paused
Immediate Protection

Enable this feature as soon as you add Fren One to your community for instant protection against spam bots and scammers targeting new member slots.

Actions​

The action(s) to be taken when a user's message is detected as spam.

You can combine multiple actions for comprehensive spam response:

Available actions:

  • Mute: Temporarily restrict the user from sending messages
  • Kick: Remove the user from the group (can rejoin later)
  • Ban: Permanently remove the user from the group
  • Delete message: Remove the spam message from the chat

Multiple Action Combinations​

Select multiple actions to create layered responses:

Example combinations:

  • Delete message + Mute: Remove spam and timeout the user
  • Delete message + Kick: Remove spam and remove user (temporary)
  • Delete message + Ban: Remove spam and permanently ban spammer
  • Mute only: Keep message visible for review but silence the user
  • Delete message + Mute + Ban: Maximum protection (remove message, restrict, then ban)
Recommended Configuration

For most communities, Delete message + Mute or Delete message + Ban provides the best balance between protection and preventing false positives from affecting legitimate users.

Duration (for Mute action)​

If Mute is selected, specify the duration of the restriction.

Common durations:

  • 1 minute: Brief timeout (warning)
  • 5 minutes: Short restriction
  • 1 hour: Standard spam timeout
  • 1 day: Extended restriction
  • 7 days: Week-long ban
  • Permanent: Indefinite mute (until manually unmuted)
Smart Duration Strategy

Start with shorter durations (1 hour) and increase for repeat offenders. Very short durations (1 minute) can serve as automated warnings.

Troubleshooting

Legitimate users being flagged?

  • Review what triggered the detection
  • Check if user truly meets "new user" criteria
  • Consider using Mute instead of Ban for first offenses
  • Add user to User Whitelist if trusted
  • Adjust "new user" definition in Filters

Spam getting through?

  • Verify Spam Detection is enabled (🟢)
  • Check bot has necessary permissions (delete, mute, ban)
  • Ensure spammers meet "new user" criteria
  • Combine with Filters for specific patterns
  • Use Captcha as additional protection
  • Report missed spam patterns (helps improve the model)

Too many false positives?

  • Switch from Ban to Mute as the action
  • Reduce duration of Mute (use 1-5 minutes)
  • Remove "Delete message" to allow admin review
  • Adjust "new user" definition to shorter timeframe
  • Enable veto-like features for community input

Actions not being applied?

  • Verify bot has admin privileges
  • Check bot has specific permissions needed (mute/kick/ban/delete)
  • Ensure the user is actually a "new user" per your Filter settings
  • Test with a clear spam-like message from new account

Want to review before action?

  • Use only Mute with very short duration (1 minute)
  • Don't enable Delete message
  • Check notifications for AI Spam events
  • Manually review and take further action

Understanding the AI Model​

What the AI Detects​

High Confidence Spam:

  • Messages with multiple external links
  • Text matching known scam templates
  • Excessive emoji usage (20+ emojis)
  • Forward spam from other groups
  • Bot-like message patterns

Moderate Confidence Spam:

  • Single promotional links from new users
  • Suspicious keyword combinations
  • Repetitive formatting patterns
  • Unusual character usage
  • Multiple messages in quick succession

Likely Legitimate:

  • Conversational messages
  • Questions and answers
  • Responses to other users
  • Simple introductions
  • Normal emoji usage

Model Improvements​

The Fren One spam detection model is continuously updated based on:

  • New spam patterns identified across all communities
  • False positive reports
  • Emerging scam techniques
  • User feedback
Model Updates

Updates to the spam detection AI model are deployed automatically. You don't need to take any action to benefit from improvements.

On this page