YouTube to deploy automatic detection for AI-generated video content
New measures will flag videos where generative AI use is detected but not disclosed, with permanent labelling applied to content created via Google tools or bearing C2PA watermarks.
YouTube has announced plans to implement an automated system designed to detect and label videos containing significant photorealistic content generated by artificial intelligence. The initiative supplements the platform’s existing policy, which mandates that creators disclose the use of realistic-looking AI tools in their uploads. Under the new framework, YouTube’s systems will analyse videos to identify signs of generative AI, applying labels where creators have failed to provide mandatory disclosures.
The labels will be displayed prominently beneath the video player for standard uploads and as overlays on YouTube Shorts. According to the platform, these tags denote the use of "photorealistic and meaningfully AI altered or generated content." The move aims to enhance transparency for users, addressing growing demand for clearer visibility regarding synthetic media within the digital ecosystem.
While creators retain the right to dispute labels they believe to be erroneous, the process includes specific conditions for permanent retention. If a video is identified as being created using Google’s own AI tools, such as Dream Screen or Veo, or if it contains C2PA watermarks—an industry standard used to flag generative AI creations—the label will remain in place permanently. In cases of disputed labels, creators may update their disclosure to resolve the issue.
The introduction of automatic detection marks a shift in how the platform manages synthetic content, moving beyond reliance solely on self-reporting. By integrating technical detection with existing disclosure requirements, YouTube seeks to close gaps where creators might omit information about the use of generative tools. The platform noted that increased transparency is a response to user requests for better visibility into the nature of content being consumed.
Details regarding the specific technical thresholds or algorithms used to define "significant photorealistic AI use" have not been provided. It remains unclear how frequently the automatic detection system will trigger compared to existing creator disclosures, or how the system will distinguish between minor edits and substantial generative alterations. The implementation is expected to roll out as part of ongoing efforts to manage the proliferation of undetected synthetic content on the service.
