YouTube announced it will auto-label videos when its systems detect significant photorealistic generative AI, rather than relying only on creator self-disclosure. Starting in May, labels will appear directly below the video player on long-form videos and overlaid on YouTube Shorts, while slightly altered or unrealistic AI content will still be tagged only in the expanded description.<\/p>\n
AI labels on the video platform have been in use for over two years, after YouTube updated its AI policies and rolled out a tool in Creator Studio that required creators to disclose videos containing AI content that could be mistaken for a real person, place, or event. Videos that obviously depicted an animated or imaginative scenario, such as a unicorn prancing through a fantastical world, did not have to be labeled.<\/p>\n
Previously, labels appeared in the expanded description, unless the video touched on sensitive topics like health or news, in which case a prominent label appeared directly on the video.<\/p>\n
YouTube says it is taking a more active role in policing content on its platform. Starting in May, YouTube will use new internal signals to help identify AI-generated content and label it accordingly. Creators should still disclose their use of AI, but if they neglect to, YouTube will label the video on their behalf.<\/p>\n
Creators whose content was misidentified can update the disclosure status on a YouTube video, but they cannot remove labels if the content was created with YouTube’s own AI tools, such as Veo or Dream Screen.<\/p>\n
Labels will be permanently attached to videos when the content contains C2PA metadata, indicating it was fully AI-generated. OpenAI recently committed to the C2PA standard, joining Nvidia, Kakao, and ElevenLabs.<\/p>\n
Now the labels will appear directly below the video player, above the description, for long-form videos, and directly overlaid on YouTube Shorts. For AI video that is only slightly altered, animated, or unrealistic, the label will appear in the expanded description only.<\/p>\n
The automatic detection update follows the expansion of YouTube’s AI deepfake detection, which now allows any adult to scan YouTube specifically for face matches, after initial tests with celebrities, public figures, politicians, and other creators. The move also follows Google’s release of Gemini Omni, a family of multimodal AI models unveiled at its Google I\/O developer conference last week that can output high-quality videos reflecting an understanding of physics, culture, history, and science.<\/p>\n
YouTube noted that AI labels will not affect how a video is recommended or its ability to monetize. Beyond labeling, the company has been investing in AI for its interactive search feature, Ask YouTube, a playlist generator for YouTube Music, AI video summaries, and other generative AI creation tools.<\/p>\n