YouTube has confirmed it quietly ran an experiment applying AI-driven enhancements to some Shorts, sparking creator complaints about consent and authenticity. Subtle changes included sharpening, unblurring, and denoising that some users noticed in skin texture, clothing wrinkles, and even ear shapes. According to BBC Future, the company did not notify uploaders or seek permission before processing affected videos.
Creators spot artifacts as YouTube details its test
Musician and YouTuber Rick Beato said his face appeared oddly altered, describing hair and skin that looked slightly artificial. Fellow creator Rhett Shull reported similar artifacts, calling the look “terrible over-sharpening” and expressing concern that it could “erode the trust” with his audience. Social media posts dating back to June compiled closeups and side-by-side comparisons questioning YouTube’s intentions.
Rene Ritchie, YouTube’s head of editorial and creator liaison, said in a post on X that the experiment used “traditional machine learning technology to unblur, denoise, and improve clarity in videos during processing,” likening it to what modern smartphones do. In a follow-up, Ritchie drew a distinction between traditional machine learning and generative AI. BBC Future reported that YouTube did not answer whether users will be able to opt out.
Debate over consent, terminology, and reality
Where AI ends and authenticity begins
Samuel Wooley, the Dietrich chair of disinformation studies at the University of Pittsburgh, told BBC Future that smartphone enhancements are user-controlled, whereas YouTube’s approach altered creator content without consent before distribution. He argued that framing the changes as “machine learning” downplays that ML is a subfield of AI, and said the distinction from generative AI was not meaningful in this context.
Jill Walker Rettberg of the University of Bergen compared analogue footprints of reality to today’s algorithmic mediation, asking what this means for our relationship with the real. BBC Future highlighted broader examples, from a controversy over apparent AI remasters of classic TV shows to Samsung’s AI “Moon” photos and Google Pixel’s AI features like Best Take and 100x generative zoom, as signs of increasingly pre-processed media.
Google’s Pixel 10 also implements new content credentials set by the industry to watermark AI-edited images, BBC Future noted. Still, Wooley said editing creator videos without their knowledge risks further undermining trust in social media content. Beato, however, said he continues to view YouTube positively, calling it a company that “changed my life.”
