The AI Death Spiral: How Smart Search Engines Are Strangling the Web

Here's how it worked until yesterday: you search for something on Google, find an interesting result, click it, land on a website, see ads, and boom — that site can keep its journalists, pay its servers, do its job. It was fragile economics but it worked. Now ChatGPT, Gemini and their friends show up saying: "Why would I let you leave my window? I'll write the answer right here inside."

Researchers like Rand Fishkin started pointing out what they call "zero-click searches" — searches where the user finds what they're looking for without ever leaving the platform. No clicks on links, no traffic to websites, no ad views. The AI summarizes, explains, tells stories — and the actual website that wrote the article? Invisible.

Caitlin Petre, a professor at Rutgers, put her finger on the wound: if traffic collapses, the economics of journalism and content creation in general risks falling apart. She's not being dramatic for kicks — it's geometric. Fewer readers = less money = fewer journalists = less quality content = less material to teach the AI. It's a real spiral, and as we descend, AI has less and less decent material to learn from.

The paradox is beautiful in a bitter way: these models were trained by reading millions of articles from paid and fact-checked journalists. Now they're sucking blood from the system that created them. It's like building a machine that robs the wallet you pulled it from.

Sure, someone will say "but this way users get answers faster, it's an improvement." Technically true. But when nobody's getting paid to write anymore, fast answers become empty answers recycled from other empty answers. AI will learn from itself, repeating endlessly.

So what's the fix? Nobody knows yet. Some publishers are asking AI companies for credit or payment. Others are putting paywalls up. But the real problem is bigger: we built a machine that eats its own supply chain.