---
url: >-
  https://docs.corbitaltech.dev/products/feedback-flow/ai-features/spam-detection.md
---
# AI Spam Detection

Automatically detect and filter spam content using AI-powered analysis.

## Overview

AI spam detection analyzes comments and feature requests to identify spam, toxic content, and policy violations.

::: tip Requirements
Requires API key from supported provider (OpenAI, Anthropic, Gemini, Azure, Perspective, or Tisane). Can also use local keyword-based detection without API keys.
:::

## Setup

1. **Settings → Spam Settings**
2. Select spam detection provider
3. Enter API key
4. Configure thresholds
5. Enable detection

## Detection Methods

### AI-Powered Detection

**Detected content:**

* Spam and promotional content
* Hate speech
* Harassment
* Violence
* Sexual content
* Profanity
* Toxicity

**Providers:**

* OpenAI Moderation API
* Anthropic Claude
* Google Gemini
* Azure Content Safety
* Google Perspective API
* Tisane AI

### Local Keyword Detection

Fallback option without API keys:

* Custom keyword lists
* Balanced or strict presets
* Multi-language support

## Configuration

### Strictness Levels

**Strict:**

* Low tolerance
* High false positives
* Maximum protection

**Medium:** (Recommended)

* Balanced approach
* Fewer false positives
* Good protection

**Lenient:**

* High tolerance
* Minimal false positives
* Light moderation

### Confidence Thresholds

Set minimum score for action (0.0 - 1.0):

* **0.8-1.0**: Very confident spam
* **0.5-0.7**: Likely spam
* **0.0-0.4**: Probably legitimate

**Auto-actions:**

* Score ≥ 0.8: Auto-block
* Score ≥ 0.6: Auto-hide, require review
* Score < 0.6: Allow, flag for review

## Automated Actions

### Auto-Block

Immediately block high-confidence spam:

* Content not posted
* User notified
* Admin log created

### Auto-Hide

Hide suspicious content pending review:

* Content hidden from public
* Visible to admins
* Can be approved/rejected

### Require Admin Approval

All flagged content requires manual review.

## Trust Score Integration

Spam detection affects user [Trust Scores](/products/feedback-flow/ai-features/trust-system):

* Confirmed spam: -10 points
* False positive (admin approved): +5 points
* Low trust users: increased scrutiny

## Reviewing Flagged Content

1. **Admin → Spam Queue**
2. Review flagged items
3. Options:
   * Approve (not spam)
   * Confirm spam (delete)
   * Ban user (severe cases)

## Performance Monitoring

Track spam detection:

* False positive rate
* Total spam blocked
* User impact
* API usage and costs

## Best Practices

* Start with medium strictness
* Monitor false positives
* Adjust thresholds based on results
* Combine with Trust Score system
* Review spam queue regularly

## Next Steps

* [Content Moderation](/products/feedback-flow/features/comments/moderation) - Manual moderation
* [Trust Score System](/products/feedback-flow/ai-features/trust-system) - User reputation