YouTube is experimenting with automatically improving some videos in its Shorts section using machine learning technology without informing content creators, the BBC and The Atlantic reported, citing a statement from the company.

According to YouTube, the experiment involves a limited number of short videos. The company uses “traditional machine learning technology” to remove noise and improve clarity in the videos. The exact scope of the test is not disclosed.

YouTube representatives said they take into account feedback from creators and viewers, but did not answer a question about the possibility of creators opting out of AI processing of their content.

The first user complaints appeared on social networks back in June 2025, but the company did not comment on the situation until recently.

Content creators react

Authors and viewers note various artifacts in the processed videos: unnatural-looking skin, strange hair changes, overly pronounced folds in clothing, harsh shadows, and a general feeling of artificiality.

Blogger Rick Beato (5 million subscribers) noticed that in one of his videos his face looked made up. The creator of the channel The Mr. Bravo Show complained about the loss of the 1980s aesthetic he intentionally recreated.

Rhett Shull (747 thousand subscribers) dedicated a separate video to the changes, comparing the result of the processing to “cheap deepfakes.” He noted that the authors are able to independently improve the quality of their videos if necessary.

Potential consequences

According to the BBC, while the changes are not always noticeable without direct comparison with the original, the use of neural networks without the consent of the authors could increase general mistrust of the authenticity of online content.

The creators also express concerns that viewers may mistake the automatically processed videos for AI-generated content, which could undermine trust in their work.