YouTube secretly touches up creators' videos with AI

Rick Beato thought he was losing his mind.

The music YouTuber, who commands an audience of over five million subscribers, was reviewing a recent upload when something looked off. His hair looked strange, his ears seemed slightly warped, and his face looked like someone had slapped makeup on him.

He wasn't imagining it. YouTube has secretly been using AI to retouch creators' videos without telling them or asking for their permission.

The secret makeover

Beato's friend and fellow music creator Rhett Shull noticed the exact same glitchy smoothing on his own channel. Shull posted a breakdown that racked up over 500,000 views, pointing out that excessive sharpening gave his clips an unmistakable, unwanted AI sheen. Online complaints about the issue actually date back to June.

Following months of speculation, YouTube finally fessed up. Rene Ritchie, the company's head of editorial and creator liaison, confirmed on X that YouTube is running an experiment on select YouTube Shorts.

According to Ritchie, the platform uses traditional machine learning to unblur, denoise, and sharpen videos during processing, arguing it works similarly to the onboard processing inside a modern smartphone. He later took pains to distinguish machine learning from generative AI.

Experts pushed back immediately. Samuel Woolley, Dietrich chair of disinformation studies at the University of Pittsburgh, called YouTube's framing a misdirection, noting that machine learning is literally artificial intelligence. When the BBC asked YouTube if creators would ever get a toggle to opt out of the processing, the company stayed silent.

A slow drift from reality

AI quietly rewriting what we see isn't entirely new. In 2023, Samsung caught heat for aggressively painting details onto photos of the Moon. Google's Pixel lineup now ships with features like Best Take, which swaps facial expressions across group shots to invent moments that never happened. Even Netflix sparked outrage in March 2025 after AI remasters of 1980s sitcoms turned backgrounds and faces into distorted mush.

Google has attempted to address transparency by baking C2PA content credentials into the Pixel 10 camera to flag AI modifications. Yet its subsidiary, YouTube, is actively altering finished user footage on the back end with zero warning.

Shull argued the meddling misrepresents his voice and threatens to erode trust with his audience. If he wanted his footage aggressively sharpened, he would have done it in the editing bay himself.

Why it matters

We think YouTube fundamentally misunderstood its mandate here. Improving playback quality is a standard platform goal, but modifying a creator's physical appearance and creative output without consent crosses a dangerous line.

For creators, this isn't just about an ugly filter. Trust is the primary currency of internet video. In an era where viewers are hyper-sensitive to synthetic media, making real humans look subtly AI-generated actively damages creator credibility. Nobody wants their audience asking why their ears look melted or whether their review was generated by a bot.

Technically, YouTube's move is clever optimization: if you can upscale low-bitrate Shorts on the server side, you save bandwidth and boost visual pop against TikTok. But the execution is deeply flawed. On a smartphone, users can turn off scene optimization and beauty modes. By forcing this upstream without an opt-out toggle, YouTube treated creator content as raw training data rather than finished artistic work. We believe platforms must draw a hard line: quality enhancements should never alter the subject's likeness, and creators must always hold the kill switch.