YouTube to begin automatically labeling AI videos
Ars Technica AI · Ryan Whitwam · 2026-05-27
YouTube will begin automatically detecting and labeling AI-generated videos using internal signals starting May 2026, ending its sole reliance on voluntary creator disclosure.
Appears in
Extraction
Topics: ai-video-detectioncontent-labelingplatform-policysynthetic-media
Claims
- YouTube will use new internal detection signals to automatically flag videos with significant photorealistic AI use, starting May 2026.
- YouTube's 2024 AI labeling system relied entirely on voluntary creator disclosure, giving uploaders no incentive to be honest.
- AI video generation tools like Seedance, Runway, and Google's Veo have improved realism to the point where viewers can no longer reliably distinguish AI content from real footage.
- Creators remain required to disclose AI use at upload time, but automated detection now supplements that requirement.
- Google's own AI content creation tools, including its Omni model, contribute to the growing challenge of distinguishing real from AI-generated video.
Key quotes
AI content creation tools like Google's new Omni model threaten to make reality even harder to discern from AI fantasy.
Starting this month, YouTube will use 'new internal signals' to flag AI content. This will apparently apply to videos that show 'significant photorealistic AI use.'
uploaders didn't have any incentive to be honest about that before.