Overview
KyberGate uses AI to monitor and classify YouTube videos that students watch on managed devices. Each video is automatically categorized as educational or non-educational, and flagged content (violence, inappropriate language, non-educational time-wasting) is reported to administrators.
This helps schools allow YouTube as a learning tool while maintaining visibility into what students actually watch.
Prerequisites
- KyberGate organization with YouTube monitoring enabled
- Students on managed devices with KyberGate filtering active
- Admin access to configure classification settings
How It Works
When a student watches a YouTube video on a managed device, KyberGate captures:
- The video URL and video ID
- Video title and channel name
- Duration watched
- Timestamp and device information
The AI classification engine then analyzes the video metadata, title, description, and channel context to assign a category:
- Educational — lectures, tutorials, documentaries, academic content, instructional videos
- Non-Educational — entertainment, music videos, vlogs, gaming, memes, unrelated content
- Flagged — content containing violence, explicit language, self-harm references, or other concerning material
Classification happens in near real-time. Results appear in the dashboard within minutes of viewing.
Steps
Step 1: Enable YouTube Monitoring
Navigate to Settings → Monitoring → YouTube Classification in the KyberGate dashboard. Toggle the feature to ON.
Step 2: Configure Classification Sensitivity
Choose your classification sensitivity level:
- Standard — flags clearly non-educational and concerning content
- Strict — also flags borderline content (e.g., educational channels with clickbait titles, music used for studying)
- Custom — define your own rules for what counts as educational based on channel lists, keywords, or categories
Step 3: Set Up Alerts (Optional)
Under Alerts, configure who gets notified when content is flagged:
- Email alerts to IT admins for flagged content
- Daily digest of non-educational viewing by class or grade
- Real-time alerts for safety-critical flags (violence, self-harm)
Step 4: Review Reports
Go to Reports → YouTube Activity to see:
- Total videos watched across the organization
- Breakdown by classification (educational vs. non-educational)
- Most-watched channels and videos
- Flagged content requiring review
- Per-student YouTube activity
Step 5: Take Action on Flagged Content
For flagged videos, you can:
- Dismiss — mark as reviewed and not a concern
- Block Channel — add the channel's domain to your blocked list
- Block Video — block the specific video URL
- Contact Student/Parent — if safety-critical content was viewed
Tips
- YouTube classification works even when YouTube is allowed — you do not need to block YouTube to monitor it
- Use the "Top Non-Educational Videos" report to identify trending distractions and proactively block them
- Classification accuracy improves over time as the AI learns from your organization's review decisions
- Consider allowing YouTube in Relaxed mode but reviewing the daily digest to ensure it is being used productively
- Pair YouTube monitoring with Bell Schedule Enforcement: allow YouTube during research periods, block during testing
- The classification does not record or store video content — only metadata (title, channel, URL, duration)
FAQ
Q: Does KyberGate watch the actual video content? A: No. Classification is based on video metadata (title, description, channel, tags) and YouTube's own category data. KyberGate does not stream, download, or analyze video frames.
Q: What about YouTube Shorts? A: Yes, YouTube Shorts are classified the same way as regular videos. The AI analyzes Shorts metadata just like full-length videos.
Q: Can I whitelist specific YouTube channels as always educational? A: Yes. In the YouTube Classification settings, add channels to the Approved Channels list. Videos from these channels are automatically classified as educational and are never flagged.
Q: Does this work with YouTube in restricted mode? A: Yes. KyberGate's classification works independently of YouTube's built-in Restricted Mode. You can use both for layered protection.
Q: How accurate is the AI classification? A: The classification engine has approximately 94% accuracy for distinguishing educational from non-educational content. Flagged content is always surfaced for human review — the AI provides a recommendation, not a final decision.