YouTube is strengthening the identification and management of AI-generated videos and no longer relies solely on voluntary disclosure by creators. The company recently announced that its internal system will automatically detect and label videos using "significant immersive AI" technology from May, including long videos and short videos, and the position and prominence of relevant tags will also be simultaneously improved.
According to information released by YouTube, the platform has been using AI content labels for more than two years. Previously, it mainly required creators to declare through tools in Creator Studio: if the video contains AI content that could be mistaken for real people, real places, or real events, it must be disclosed; while content that is obviously fantasy, cartoon, or unrealistically confusing, such as unicorns running in a fictional world, is exempt from labeling. YouTube stated that the current AI labeling policy itself has not changed, but the platform will shift from "requiring creators to self-report" to a more active governance model of "platform active identification + creator cooperation".
This move comes on the heels of Google’s release of the Gemini Omni multimodal model at its I/O developer conference. Gemini Omni can generate high-quality videos that understand complex backgrounds such as physics, culture, history, and science, raising concerns about the further proliferation of immersive AI videos. YouTube said it will use new internal signals to identify AI-generated content and automatically label it. Even if the creator forgets or refuses to disclose it, the system will label it on his behalf.
In terms of labeling correction, if creators believe that their content has been misjudged as an AI video, they can update the disclosure status in the video. However, if the content is generated by YouTube's own AI tools (such as Veo or Dream Screen), the creator does not have the right to remove the AI label added by the platform. Additionally, once a video contains C2PA metadata and appears to be fully AI-generated, the associated tags will be permanently attached to the video. Currently, OpenAI has announced its support for the C2PA standard, and organizations such as NVIDIA, Kakao, and Eleven Labs have also joined this camp to promote the traceability of AI content.
Prior to the launch of automatic detection and annotation, YouTube has expanded the scope of its AI deepfake detection tools. The platform’s face-matching scanning tool, which was originally only available for testing to celebrities, public figures, politicians, and some creators, can now be used by any adult to scan whether their image on YouTube has been artificially synthesized or impersonated by AI. On this basis, YouTube believes that it is necessary to simultaneously expand the scope and intensity of automatic labeling of AI content.
YouTube will also adjust the placement of labels to make them more visible and consistent. Previously, AI labels were usually hidden in the expanded description column, and would only be prominently displayed at the top of the video screen when the video involved sensitive issues such as health and news. After the revision, for long videos, the AI tag will be fixedly displayed directly below the player and above the video description text; for Shorts videos, it will be directly superimposed on the video. The platform stated that this design is to allow users to identify immersive content that has been modified by AI or generated by AI at a glance.
For videos that are only lightly AI-modified, animated, or clearly unrealistic (such as the aforementioned fantasy unicorn scene), YouTube will still place the label in the expansion description rather than under the player or overlaying it to avoid unnecessary disruption to the viewing experience. It is worth noting that YouTube emphasizes that the AI tag itself will not affect the video recommendation algorithm, nor will it weaken its monetization ability, that is, works marked as AI content can still participate in distribution and advertising revenue normally.
While strengthening AI content management, YouTube also continues to use AI technology for product innovation and feature upgrades. The platform has previously tested and launched a number of AI features, including the interactive search function "Ask YouTube", an AI playlist generator for YouTube Music subscribers, AI automatic video summarization, and a series of generative creation tools and AI music generation tools for Shorts and creators. For YouTube, how to strike a balance between using AI to promote product development and preventing the risk of AI abuse is becoming one of the key issues in its future strategy.