21% of YouTube Shorts are AI-made as Snap, LinkedIn crack down on synthetic sludge

The internet’s synthetic sludge problem just hit a tipping point. Snap confirmed this month it’s pulling wholly AI-generated videos from its Spotlight recommendations, a move designed to protect what the company calls “authentic creativity from real people.” Videos edited with Snapchat’s own AI tools survive the purge but will now carry transparency labels. The timing is awkward: Snap boasted that Spotlight contributors grew over 120 percent, yet it won’t say how much of that surge was fueled by the very machine-made content it’s now booting off the platform.

Snap isn’t acting in a vacuum. A Kapwing study found that 21 percent of YouTube Shorts are already AI-generated, and anyone scrolling Instagram or Facebook can see the same deluge. The trend has birthed some spectacular failures along the way — OpenAI’s social video app Sora landed with a thud, proving that simply stuffing generative AI into a social feed doesn’t make it compelling. Users can smell inauthenticity from a mile away, and platforms are finally acknowledging that the race to fill feeds with algorithmically generated filler erodes the very thing that makes social media sticky: the sense that a real person is on the other side.

LinkedIn, of all places, has become ground zero for AI-generated junk — think bland, vaguely inspirational posts that read like a Hallmark card written by a Markov chain. The professional network rolled out a dedicated “AI slop” reporting button, but it’s a half-measure that creates its own headaches. The tool is vulnerable to misuse, and it puts the burden on users who need to know how to spot synthetic content reliably — a skill most people simply don’t have. It’s like handing someone a spam filter and saying “good luck figuring out what’s spam.”

What makes this moment different from previous content moderation panics is the sheer volume. Previous waves of low-quality content — SEO farms, engagement bait, copy-paste memes — were constrained by human effort. AI removes that bottleneck. A single prompt can generate thousands of posts, and the infrastructure for distributing them already exists. The platforms’ current countermeasures feel like using a bucket to bail out a sinking ship. Transparency labels are a start, but they don’t answer the harder question: when the cost of generating content approaches zero, what does “authentic creativity” even mean anymore, and who gets to decide?

Look, I’ve been watching the content moderation pendulum swing for years, and this feels different. The previous playbook assumed bad content came from bad actors you could identify and ban. AI-generated sludge comes from everywhere — hobbyists, marketers, teenagers goofing around. The platforms are trying to draw a line between “AI-assisted” and “AI-generated,” as if the distinction will hold. It won’t. The real fight isn’t about detection tools or reporting buttons. It’s about whether these platforms can rebuild a recommendation system that rewards human effort when the cheapest, fastest content on earth is also the most abundant.