Glue-pizza recipes have colonized Google Search, sham biographies are stuffing Amazon product listings, and "Shrimp Jesus" images are flooding Facebook feeds. This low-quality content — collectively known as "AI slop" — has finally pushed the very companies that created it to act. According to The New York Times, a "generative garbage cleanup" is underway across Silicon Valley: Spotify removed 75 million bulk-uploaded "slop" tracks last year; YouTube cleaned up 130,000 low-quality AI channels in six months, while researchers wiped out 50,000 coordinated fake-account clusters; LinkedIn declared that "combating AI slop is everyone's top priority" and shipped a reporting button; TikTok has auto-labeled more than 3 billion AI-generated videos.
The problem has become impossible to ignore: Merriam-Webster named "slop" its 2025 word of the year. On French streaming service Deezer, nearly half of tracks uploaded each day are AI-generated — many never heard even once. Over a third of uploads to Apple Music are AI content. Experiments show that more than 20% of the first short videos recommended to new YouTube users are AI slop — and one top slop channel was earning an estimated $4.25 million a year.
- Original content is being "fast-fashion-ified": cheap, disposable content squeezes human creativity the way fast fashion squeezes haute couture
- Slop is feeding back into AI: synthetic content gets scraped for training, creating a vicious cycle of garbage producing garbage
- Monetary incentives are twisted: a single slop channel earning millions means the platform's payout system is effectively subsidizing dumping
-
Trust is nearing a breaking point: as Substack's CEO put it, "it's getting harder and harder to tell what's real on the internet"
-
Meta: signed a European pledge to make AI content identifiable through detection and labels; yet its new image generator auto-ingested public Instagram accounts as reference data and was pulled within three days after user backlash
- Pinterest / TikTok: giving users control — letting them cap the amount of AI content they see
- X: with nearly half of 50+ word posts AI-generated, the platform banned apps that pay bots to post
- Substack: partnered with AI-detection startup Pangram on a detector, already drawing complaints that it flags human-written posts as AI
The irony is hard to miss: many of the companies now sweeping up the trash are the same ones that made it — their own AI tools spawned the slop, and now AI is being recruited to clean it up.
As a regular user, you don't have to wait for platforms to save you. A few simple checks filter out most AI slop: look for weird details (extra fingers, nonsensical text), and beware of clickbait headlines with no author and high-volume templated posting. Your reports genuinely matter now — LinkedIn, YouTube and others have opened dedicated channels, and user feedback is becoming a key signal in the cleanup.
For creators, there's a silver lining: slop may be drowning the feed, but precisely because cheap content is being devalued, genuinely original work is becoming rarer — and more valuable.
Columbia University researcher Camille François puts it plainly: "We can't simply define it and then delete it." AI slop isn't a category of content — it's a systemic problem of the information ecosystem. As long as generation costs approach zero and traffic monetization persists, the garbage keeps coming. Platforms that only treat symptoms will watch users vote with their feet.
It's also a wake-up call for the AI industry: AI's real edge should be amplifying human creativity — not flooding the internet with industrial wastewater.