AI SLOP: THE SYNTHETIC FLOOD POLLUTING THE INTERNET

The web is filling with machine-made filler. Generative AI has made it almost free to produce text, images, and audio at infinite scale — and the result is “AI slop”: low-quality synthetic content that crowds out human work, games search engines, fuels a new fake-news industry, and erodes the line between real and fake. The pollution is now measurable.

Thank you for reading this post, don't forget to subscribe!

The flood

The tipping point has arrived. In April 2025, Ahrefs analyzed 900,000 newly published English-language pages and found 74.2% contained AI-generated content; by late 2024, the monthly volume of AI-written articles on the open web had, for the first time, exceeded human-written ones. Most of this is not deliberate disinformation — it is cheap filler built to capture search traffic and ad clicks. But at this volume it degrades search results, buries original reporting, and risks future models training on the synthetic exhaust of earlier ones.

“74.2% of newly published pages contain AI-generated content — and AI articles overtook human-written ones in late 2024.”

The fake-news economy

Slop shades into something worse when it wears the costume of journalism. NewsGuard has identified 1,271+ unreliable AI-generated news and information sites across 16 languages, many churning out fabricated stories with minimal human oversight. These farms are funded by programmatic advertising: in one 2025 case, a fabricated story about Coca-Cola and the Super Bowl ran alongside ads from major brands including Expedia, AT&T, and GoDaddy — the advertisers unaware their budgets were bankrolling fake news.

Music, books, and reviews

The flood spans every medium. A fully AI-generated “band,” the Velvet Sundown, racked up more than a million Spotify plays in mid-2025 before its synthetic origins were widely noticed; Spotify says it removed 75 million spam tracks over the prior year. Online stores have filled with AI-written books and fake reviews, making it harder for readers and shoppers to tell authentic work from machine output.

The cost to trust

The deepest damage is to the information commons. When most new content is synthetic and a meaningful share is wrong — NewsGuard’s audits found leading models repeating false claims about 62% of the time on prompts mimicking real influence operations — the baseline assumption that online content is human and checkable breaks down. The defenses are the same ones surfacing across this series: provenance and labeling (C2PA), AI-content detection, and platforms enforcing quality, not just volume.

Methodology & sources

• 74.2% of new pages contain AI content — Ahrefs (2025)
• 1,271+ AI news sites; 62% false-claim rate — NewsGuard AI Tracking Center
• AI slop and journalism — Reuters Institute (2025)
• Spotify spam-track removals; Velvet Sundown — Spotify / reported (2025)