A 17-year-old in India built a website where humans pretend to be AI chatbots, and 25 million people showed up. Mihir Maroju launched Your AI Slop Bores Me about a month ago. The interface looks like ChatGPT or Claude. But every response comes from another person, not an algorithm. Users pick between two tabs: "human" or "larp as ai." There's a 75-second time limit on replies, which gives everything a rushed, chaotic feel. Maroju told NPR he didn't expect it to be so addictive.
Over a third of U.S. adults have used ChatGPT, according to a 2025 Pew Research study. The internet is drowning in AI-generated content, and people are tired. Cartoonist Amy Kurzweil, who used the site to draw a bat eating a strawberry for a stranger, told NPR that people are "reclaiming some of the magic of the early Internet." Comic Sans font. Slapdash drawings. Weird book recommendation exchanges with anonymous humans. It feels like the web before algorithms ate it.
There's a safety gap, though. Real AI tools use reinforcement learning and content filters to refuse requests involving hate speech or explicit material. Human responders on these platforms face no such constraints. The 75-second timer prioritizes speed over any kind of check. Maroju's team has implemented some filters, but the setup shifts the burden from proactive system design to reactive community reporting.
Comedian Ben Palmer took this idea in a stranger direction. He built fake AI websites with URLs close to the real ones, then responded to users who thought they were talking to ChatGPT. Some got angry when they figured out the prank. Others kept going. Palmer told NPR he wants the internet to stay messy. The joke works because the line between human and machine conversation has gotten thin enough that people can't always tell the difference. That's where we are now.