A researcher invented a fake eye disease called “bixonimania” and uploaded two bogus papers about it to an academic server. The papers included acknowledgments to the “Starfleet Academy,” funding from a character in The Simpsons, and the “University of the Fellowship of the Ring.” In the middle of the text, it was explicitly stated that everything was fictional.

Nevertheless, for several weeks, major AI systems treated the disease as real: Google Gemini claimed it was caused by blue light, Perplexity reported a prevalence of one case per 90,000 people, and ChatGPT even advised users on matching symptoms.

The fake study was eventually cited in a peer-reviewed journal, which later retracted the issue after intervention by Nature.

Neither AI systems nor human researchers initially detected the hoax—highlighting a growing problem: people are citing AI-generated references without verifying their content.

Meanwhile, the U.S. FDA is already using AI to evaluate drugs, the CEO of a New York hospital is considering replacing radiologists with algorithms, and ChatGPT Health is being launched to consult patients.

@science