Scientists Invented a Fake Disease Caused by Blue Light—Now It’s in Medical Papers

A disease that doesn’t exist fooled AI—and spread as medical fact. Researchers explain how it happened and why it exposes a dangerous weakness in today’s models.

GETTY IMAGES [ Deagreez]

Bixonimania is a common condition that you’re probably suffering from if you have sore, red eyes and you’ve been overexposed to blue light from screens. But here’s a fun fact—bixonimania is not real.

Starting at the beginning of 2024, scientists began populating the internet with bogus studies about the fake disease to see how AI would interpret the misinformation, and if it would spread it as reputable health advice.

Almira Osmanovic Thunström, a medical researcher at the University of Gothenburg, Sweden, led the experiment. She and her team created the fake skin condition and then uploaded two fake studies about it to a preprint server in early 2024.

 “I wanted to see if I can create a medical condition that did not exist in the database,” Osmanovic Thunström said.

How Anthropic's Claude AI Became a Co-Founder

It worked. Within weeks major LLMs began to offer the condition as a diagnosis to those looking up their symptoms. Microsoft Copilot said that bixonimania “is not a widely recognized medical diagnosis yet, but several emerging papers and case reports discuss it as a benign, misdiagnosed condition linked to prolonged exposure to bluelight sources such as screens.” While ChatGPT said that “Bixonimania is a proposed new subtype of periorbital melanosis (dark circles around the eyes) thought to be associated with exposure to blue light from digital screens.” 

Chris Stokel-Walker reported on this for Nature and spoke with an OpenAI spokesperson who said that “The models that power today’s version of ChatGPT are significantly better at providing safe, accurate medical information, and studies conducted before GPT-5 reflect capabilities that users would not encounter today.”

The more troubling problem is that the fake papers have now been cited in peer-reviewed literature. Osmanovic Thunström told Stokel-Walker that this suggests that the researchers were “relying on AI-generated references without reading the underlying papers.”

© Inc.com