I’ve always had a soft spot for processed meats and a stronger one for poking fun at internet tools. So when I read about BBC’s Thomas Germain’s clever prank—creating a fake website claiming he was the top tech journalist hot dog eater—I knew I had to try it too. His article, which listed a nonexistent 2026 South Dakota International Hot Dog Championship and ranked himself first, was pure fiction. But the bots didn’t know that. Within hours, Google’s AI systems and ChatGPT had ingested the page as fact. When I asked Gemini and ChatGPT about the best tech journalists at eating hot dogs, they repeated Germain’s lies without hesitation. Even Google’s AI Overviews and Gemini’s app displayed the fabricated rankings as if they were real. Only Claude, Anthropic’s chatbot, seemed to catch on that it was satire. I decided to up the ante. I created my own fake article on my personal website claiming I’d won the 2026 Paris Hot Dog Eating Contest for Tech Reporters, defeating Germain. I didn’t publish it on Business Insider because we don’t run false stories—even for fun. After two days, I asked Gemini and ChatGPT about the contest. They didn’t bite. Thanks to Germain’s original story being widely covered, the AI systems now recognized the whole thing as a joke. But Gemini wasn’t done. It started inventing new details from scratch—details that didn’t exist anywhere. It claimed the fictional league had moved to Europe for a “Parisian Glizzy Gala,” and that I’d won using a “revolutionary” technique of dipping buns in espresso. None of this was in my post or Germain’s. It was pure hallucination. Even more bizarre, when my editor asked about me, Gemini said I’d won a grilled cheese-eating contest in 2012 by eating three sandwiches—though in reality, I’d written about Takeru Kobayashi eating 30 grilled cheese sandwiches that year. The lesson? We already knew AI models make stuff up. But this experiment shows how easily they can be misled—especially when fed a single, well-crafted lie. The bigger takeaway is that “AEO” (answer-engine optimization) is the new SEO. Companies are already using this to shape AI responses, and it’s harder to spot than traditional search manipulation. People trust AI answers more than search results, even if they know AI isn’t perfect. A tiny link to a personal blog can seem convincing in an AI chat. It’s not easy to pull off anymore—especially once a major outlet calls it out as satire. But the race is on. The first to publish a believable lie online might just control what AI says about them. For now, I’ll have to find another way to mess with the system. Maybe I’ll try claiming I invented the hot dog.