YouTube Enhances AI Content Labeling and Detection
Google’s YouTube is rolling out significant changes to how it identifies and displays AI-generated content, featuring new auto-discovery functions and more prominent labels for viewers.
According to a report, YouTube is upgrading its labeling system for AI-created videos, introducing automatic detection mechanisms and making disclosures more noticeable. These policy changes build on initiatives first established in 2024, which required creators to manually indicate their use of AI tools.
The platform detailed these updates in a blog post, explaining that the goal is to boost transparency while still giving creators control. Beginning this week, YouTube will implement a new internal system designed to automatically recognize videos that use photorealistic AI technology.
While creators must still disclose any use of realistic AI manually, YouTube will now automatically apply labels to content identified by its detection system as having significant photorealistic AI usage, even if creators don’t disclose it themselves. This automatic labeling intends to help viewers quickly spot content that might look genuine but is actually AI-generated or significantly altered.
If creators think their videos have been wrongly labeled as AI-generated, they can adjust their publishing status using YouTube Studio tools. However, YouTube indicates that some labels will remain permanent. This is especially true for content produced using YouTube’s own AI tools, such as Veo and Dream Screen, or content tagged with C2PA metadata indicating full AI creation.
In addition to detection features, YouTube is also altering the placement of AI content labels to enhance visibility. Previously, disclosures were tucked away in the video’s expanded description section. The new updates will ensure that labels for photorealistic and AI-modified content are much more prominent.
Interestingly, a report noted that about 20% of new videos on YouTube are generated by AI, a number that includes many low-quality videos aimed primarily at children.
A study by an editing company found that over 20% of videos recommended to new users on YouTube fall into what’s termed “AI slop,” referring to low-quality content crafted by AI solely for view monetization. This study analyzed 15,000 popular YouTube channels globally and discovered 278 channels dedicated exclusively to this type of content.
These channels together boast over 63 billion views and 221 million subscribers, raking in around $117 million annually. When researchers created a new account, they found that 104 of the first 500 recommended videos were classified as AI slop, and roughly a third of the recommended videos were deemed to have “brain rot” characteristics, which includes both AI slop and other low-quality attention-grabbing content.
