YouTube’s New Transparency Policy on AI-Generated Content
YouTube has announced a significant change in its transparency policy. Starting this month, the platform will move away from relying solely on content creators to label AI-generated videos. Instead, YouTube will implement its own detection systems to identify and label videos produced using artificial intelligence. This change aims to enhance visibility for both regular videos and Shorts by making labels more apparent to viewers.
What Exactly Changes?
Traditionally, the onus was on content creators to disclose AI use. Beginning in 2024, YouTube will mandate creators to identify when they utilize realistic AI that can mislead viewers regarding real people, places, or events. Unfortunately, many creators lacked strong incentives to be honest about AI use.
Now, according to YouTube’s official statements, if AI is used significantly—despite creators not indicating so—the platform will automatically apply an AI label to the video.
More Visible AI Labels
A crucial aspect of this new policy is the redesigned placement of AI labels. Previously, AI-related information was often buried in the extended video description. Moving forward, for longer videos, the label will be prominently displayed just below the player, above the description. In the case of Shorts, the label will be overlaid directly on the video itself, making it easier for viewers to quickly identify AI-generated content. YouTube aims to provide viewers with context “at a glance.”
What the System Does Not Cover
The new AI tags will specifically apply to photorealistic or heavily modified AI content. Videos that are animated, clearly unrealistic, or feature only minor alterations will still carry disclaimers but will remain in the extended description. This differentiation is essential as it separates potentially misleading AI usage from more benign entertainment forms created using AI.
Detecting AI in Videos
Although YouTube has been vague about the specific methods for detecting AI-generated content, they have noted two definitive criteria for automatic labeling: videos containing C2PA metadata, which marks them as entirely AI-generated, and videos produced using Google’s proprietary tools, such as Veo or Dream Screen. If a user believes their video has been incorrectly labeled, they’ll have the opportunity to correct it in YouTube Studio, except in the two aforementioned cases, where the label is permanent.
A Reasonable Decision Amid Challenges
This change comes on the heels of Google’s unveiling of the Gemini Omni model family, capable of generating increasingly realistic AI videos. YouTube and various social media platforms have been grappling with an influx of ‘AI Slop,’ particularly in Shorts, and the company asserts improvements to its deepfake detection technology, which now allows users to search for matches of their facial features on the platform.
No Impact on Monetization or Recommendations
Interestingly, YouTube has stated that a video carrying an AI tag will not detrimentally affect its recommendation algorithm or monetization potential. The platform emphasizes that its goal is “to balance transparency with creator control,” rather than penalizing those who utilize AI in their videos.
What Lies Ahead
It remains to be seen how effective the automatic detection will be and how many videos may bypass the labeling system. YouTube has acknowledged that some AI-driven content may still evade detection. Given the previous inconsistencies in its labeling practices, stakeholders will be watching closely to see how effectively this new initiative holds up in practice.
Cover image | YouTube
In Xataka | Nvidia CEO Jensen Huang: “People talk about AI eliminating jobs. It’s complete nonsense.”