YouTube is rolling out automatic detection for AI-generated videos. If creators don't disclose using generative AI tools and the platform detects "significant photorealistic AI use," a label gets applied automatically.
This is different from YouTube's existing disclosure requirements, where creators self-report AI usage. That system relied on honesty. The new system doesn't.
The detection targets photorealistic content that's been meaningfully altered or generated with AI. That includes videos made with Google's own tools like Dream Screen and Veo, as well as content bearing C2PA watermarks, an industry standard for marking AI creations.
If YouTube detects Google AI tools or C2PA markers, the label becomes permanent. Creators can't remove it. For other detected AI content without those specific signatures, creators can update their disclosure settings to reflect actual AI usage, and the label gets removed.
The dispute process is straightforward: if you think YouTube mislabeled your content, update your AI disclosure. If you actually used AI, the label stays. If you didn't, it goes away.
Label placement has also changed. For regular videos, it appears directly beneath the video player where viewers can't miss it. For Shorts, it shows as an overlay on the video itself. This is much more visible than the previous implementation, which buried disclosures in description text.
What's interesting is what this reveals about AI detection technology. YouTube clearly has enough confidence in their detection systems to automatically apply labels that creators can't remove. That suggests the technology for identifying AI-generated content is getting reliable, at least for photorealistic material.
The focus on photorealistic content makes sense. Stylized AI art or obvious digital effects aren't what misleads viewers. Deepfakes and synthetic media that looks real are the problem. YouTube is targeting the specific use case that raises trust issues.
