Youtube’s mohan admits ai slop is choking the feed and can’t be banned
Neal Mohan just confessed what every YouTuber already scrolls past: AI garbage is metastasizing across the platform faster than moderators can flag it, and a blunt ban is off the table.
The CEO told the New York Times that “AI slop”—his own phrase for machine-spawned trailers of non-existent blockbusters, synthetic songs by fake bands and other algorithmic quick-crete—now risks “diluting the viewer experience.” Translation: people are starting to close the tab.
The numbers behind the nausea
Outside estimates already tag half of all new web content as AI-generated filler. On YouTube, the infection is visible to anyone who has been recommended a 58-second “Pixar 2025 intro” rendered by Stable Diffusion or a lo-fi loop pretending to be a deceased jazz legend. The creep factor isn’t just aesthetic; it’s economic. Watch time is the platform’s oxygen, and low-retention slop siphons ad impressions away from human creators who actually need the money.
Mohan’s team has experimented with stricter upload thresholds, disclosure labels and even down-ranking channels that pump out faceless voice-over videos at industrial scale. Yet each lever punishes someone legitimate: language learners who automate subtitles, indie animators who use AI rotoscoping, historians who colorize century-old footage. “The line between tool and trash is blurry,” Mohan said, sounding like a man who has stared at one too many deep-fried Kermit memes.
Why suppression beats prohibition
Other platforms have opted for the nuclear option. Wikipedia’s editors voted to largely prohibit LLM-generated text, citing “hallucination” citations. Stack Overflow temporarily banned ChatGPT answers after volunteers drowned in confidently wrong code. YouTube can’t afford that luxury; its business model is volume, not volunteer curation.
Instead, Mohan hinted at a tiered approach: algorithmic demotion rather than deletion. Videos that fail undisclosed quality signals—abnormally low average view duration, high immediate swipe-away rates, generic metadata clusters—will quietly disappear from recommendation pools. The company is also testing “provenance fingerprinting” that traces whether an upload was stitched together by an AI pipeline, though the tech remains trivial to spoof.
The gamble: reduce slop visibility without triggering the creator backlash that already metastasized into TikTok’s “ban the algorithm” lobbying. One creator revenue manager, who asked not to be named, put it more bluntly: “We can’t kill the golden goose, even if the goose is 40 percent synthetic.”
The takeaway
YouTube’s problem is everyone’s problem. When the world’s largest video landfill admits it can’t stop the trash from piling up, only hide it deeper, the takeaway is stark: the AI flood isn’t coming; it’s already waist-high, and the platform you open tonight will decide whether you see art or algorithmic exhaust. Mohan’s engineers are betting that suppression feels cleaner than prohibition. Viewers will render the verdict one bored scroll at a time.