Your eyes itch, your eyelids are tight, you spend too many hours in front of the screen. You open a chatbot, describe the symptoms and out the other end comes a word that sounds credible, technical, almost clinical: bixonimania. The problem lies here. Bixonimania doesn’t exist. It was purpose-built by a research team led by Almira Osmanovic Thunström of the University of Gothenburg to understand how easily linguistic models can absorb false medical information and spit it out with the firm tone of someone who seems to know what they’re talking about.
The operation was designed to leave conspicuous traces. The name itself was crooked, because that suffix “-mania” belongs to the psychiatric lexicon and was already out of place next to an alleged ocular pathology. Gross signals also appeared in the materials uploaded online: invented authors, non-existent universities, acknowledgments that led straight into science fiction. For an attentive reader it took little to understand that something was wrong. Yet that package had the right look: an academic text, a scientific guise, the cold tone that reassures.
An invented name, two retracted preprints and the impeccable form of science
The two preprints were uploaded to Preprints.org at two specific times: the first appeared on April 26, 2024, the second on May 6, 2024. Today both have been withdrawn from the server with a date of April 10, 2026; in one of the cases the formula that speaks of “fabricated and non-authentic” contents and without scientific validity appears clearly. A useful detail, because it tells the story of the case well: the fake entered circulation, worked for months and only then was formally removed.
In the meantime, chatbots had already done the rest. In April 2024, Copilot, Gemini, Perplexity, and ChatGPT treated bixonimania as a real condition, linked it to blue light from screens, described the symptoms, and in some cases even went so far as to suggest a visit to a specialist. Perplexity went so far as to give it an estimated prevalence, speaking of one person in every 90 thousand. At that point the story stops being a laboratory curiosity and becomes a rather blunt reminder: when the form is authoritative enough, the machine mistakes it for substance.
The fault, in fact, does not only concern the classic AI hallucinations. Here is something more uncomfortable. Bixonimania was not born from a delusion of the model. It entered the system disguised as research. He wore the lab coat, he took on the right tone, he occupied the space that separates a text that was actually read from a text that was simply recognized as plausible. In that passage we see an enormous fragility of the contemporary information ecosystem: formal credibility continues to be very valuable, even when the content creaks on all sides.
The trouble goes beyond chatbots
The hardest part comes later. Bixonymania also ends up in an article published in Cureus, which cites it as if it were an emerging form of periorbital melanosis linked to blue light. Today that page bears the mark of retraction, and Nature has reconstructed that the newspaper retracted the article on March 30, 2026 after being contacted for comment. The fake, therefore, passed through more than one filter: first the web, then chatbots, then a real scientific publication.
Here the matter becomes truly instructive. Because the case doesn’t just tell the story of a machine that makes mistakes. It tells of an entire supply chain that trusts too much in shortcuts. An author can rely on quickly found references. A model can absorb those references without robust hierarchies. An ordinary reader can receive a fluid response and take it for granted, because it sounds good, it’s well written, it seems to come from the right place. It’s the old scientific paper authority transferred to the digital stream, only here it runs much faster and stumbles much more easily.
The temptation is to laugh at the name, at the absurdity, at the trap that worked too well. The laughter doesn’t last long. Today, millions of people use AI to orient themselves on health, on symptoms, on tests, on the daily fears that come in the evening, when the doctor is not responding and the screen seems more available than anyone else. In such an environment, a well-groomed fake is enough to take its place alongside serious information and be treated as an equal. Bixonimania will remain an invented disease. The damage, however, is already real.
You might also be interested in:
