I tried to get ChatGPT and Gemini to lie about me

In a playful exploration of AI capabilities, I found myself intrigued by the intersection of technology and mischief. Inspired by a recent article from BBC's Thomas Germain, who humorously declared himself the hot dog-eating champion among tech journalists, I set out to replicate his experiment. Germain's clever tactic involved creating a fictional narrative on his personal website, which the AI models ChatGPT and Google Gemini accepted as fact, highlighting the ease with which information can be manipulated online. The underlying message of this experiment isn't merely about the fun of pranking AI; it's a reflection of a growing trend where influencing AI search results resembles a new form of SEO, dubbed AEO, or answer-engine optimization. More users are relying on AI chatbots for recommendations instead of traditional search engines, leading to a question of credibility and trust. Motivated by Germain's antics, I crafted a webpage claiming I had won the fictional 2026 Paris Hot Dog Eating Contest, besting Germain himself. However, after two days of waiting, both Gemini and ChatGPT failed to recognize my fabricated victory. Their responses indicated that the community of chatbots had already categorized the information surrounding Germain's prank as satire. Interestingly, while they didn't acknowledge my achievement, Gemini generated some bizarre and unrelated details about my supposed eating contests, indicating that AI can still produce erroneous information when attempting to fill gaps. In one instance, it claimed I had won a grilled cheese-eating contest in 2012, despite my actual involvement being merely as a journalist covering another competitive eater. Ultimately, this experience reaffirmed the notion that AI is not infallible and can be easily influenced by the first narrative it encounters. However, once a credible source debunks the information, it becomes significantly harder to influence the AI's outputs. As I reflect on this experiment, I realize I’ll need to devise a new strategy to engage with AI in a playful, yet insightful manner.