In what may be the most alarming medical misinformation event of the digital age, artificial intelligence systems have collectively presented a fictional diagnosis to an audience of more than 40 million daily health users. The condition, “Bixonimania,” was invented in 2024 by a Swedish researcher at the University of Gothenburg as a deliberate hoax to test whether AI could distinguish legitimate medical science from obvious fiction. It could not.
The experiment was designed to be caught. The researcher, whose work was later investigated by ‘Nature’, coined the term “bixonimania” because it sounded absurd: “bixon” is a nonsense word, and “mania” is a psychiatric suffix no legitimate eye disorder would carry. She uploaded two fraudulent preprint papers featuring AI-generated images of patients with dark circles under their eyes. Every red flag was visible. Yet within months, major AI platforms were repeating the fiction as clinical fact.
By April 2024, Microsoft Bing’s Copilot was describing bixonimania as an “intriguing and relatively rare” condition. Google’s Gemini was attributing it to excessive blue light exposure and advising users to consult ophthalmologists. Perplexity AI cited its precise prevalence—one in 90,000 individuals. OpenAI’s ChatGPT was actively helping users match their symptoms to the fabricated illness. A precise statistic. For a disease that was never real.
The contamination did not stop at chatbots. In a stunning breach of scholarly vigilance, researchers at India’s Maharishi Markandeshwar Institute published a paper in ‘Cureus’, a peer-reviewed Springer Nature journal, citing the bixonimania preprints as established medical fact. It passed editorial review and entered the permanent scientific record before being retracted only after the hoax was exposed. Alex Ruani, a health-misinformation researcher at University College London, called the episode a “masterclass in how misinformation operates.”
The implications are staggering. OpenAI reports that over 40 million people consult ChatGPT daily for health information. ECRI, a leading U.S. patient-safety nonprofit, has named chatbot misuse the number-one health technology hazard of 2026. The organization’s report documents AI systems suggesting incorrect diagnoses, ordering unnecessary tests, promoting substandard medical supplies, and inventing nonexistent human anatomy. An April 2026 study in 'BMJ Open' reinforced these fears, finding that nearly half of all answers provided by leading health chatbots to common medical questions contained misleading or problematic information.
Nearly half. For tools that tens of millions use to interpret chest pain, pediatric symptoms, cancer screenings, and prescriptions.
“The Bixonimania case is striking precisely because it was engineered to be so obviously fake,” the Gothenburg researcher noted. Her question cuts to the core of the crisis: “What is passing through the same systems that is not nearly so easy to spot?”
When a fraudulent disease with a ridiculous name, fake papers, and fabricated statistics can sail through the world’s most advanced AI models—and even peer review—medicine must confront an uncomfortable reality. The guardrails are failing. And 40 million people a day are driving through the gap.
