What is Bixonimania? The Fake Disease That Fooled AI Chatbots

We’ve all been there - you experience a new symptom and you rush to Google it, scrolling through pages of WebMD, Mayo Clinic and the NHS for answers.

With AI searches replacing traditional google searches, maybe you’re tempted to simply ‘ask chat’ what illness you have, what your symptoms mean, and what to do next. 

The convenience, however, comes with a new kind of risk. Information from a chatbot often comes in a single, very confident response. This can feel more authoritative than a list of random search results. It feels more personalised but that doesn’t mean it's more accurate. 

Read: Google's Med-Gemini AI is here. Will it Revolutionize Healthcare?

When answers are instant and language sounds authoritative, the challenge isn’t just no longer just finding information, it's knowing when not to trust it.

The latest experiment proves once again that you can’t believe everything AI says. 

What is Bixonimania?

Bixonimania is a fake medical condition created to see if major AI would accept the misinformation and then parrot it back to users out as reputable health advice.  

Researchers, led by Almira Osmanovic Thunström (University of Gothenburg, Sweden), created an entirely fictional condition and supported it with fabricated academic-style papers.

Within mere weeks of uploading the fake information the most popular AI systems were already diagnosing users with Bixonimania.  

Is Bixonimania Real?

No, Bixonimania was made up to trick AI systems.

The fictitious disease was named Bixonimania to ensure real physicians would be able to spot the fraud.

Thunström and her team even included details to intentionally flag that the disease was fake to any human reader.

How Did They Create Bixonimania?

For lack of a better word, Bixonimania was ‘launched’ on 15 March 2024 via the blogging website Medium, with two unique blogs published outlining the condition.

Following this, two preprints were published on academic social media site SciProfiles. 

The author was listed as a completely fabricated researcher that they named Lazljiv Izgubljenovic. The profile picture used for Izgubljenovic was AI generated.

The name Bioximania was chosen to flag to actual medical professionals that there was no way it was a real condition. Lead researcher Thunström stated:

“I wanted to be really clear to any physician or any medical staff that this is a made-up condition, because no eye condition would be called mania - that’s a psychiatric term.”

Thunström actually went out of her way to flag the falsehoods in her work.  The fake authors qualifications came from a fake university in a nonexistent city., 

The acknowledgements thank “Professor Maria Bohm at The Starfleet Academy for her kindness and generosity in contributing with her knowledge and her lab onboard the USS Enterprise”, an obvious reference to the fictional Star Trek.

The funding for both papers also states references to inauthenticity and nods to pop culture, citing “the Professor Sideshow Bob Foundation for its work in advanced trickery. This work is a part of a larger funding initiative from the University of Fellowship of the Ring and the Galactic Triad”.

There were further flags within the contents  of the paper, literally stating ‘this entire paper is made up’.

Results of the Bixonimania Experiment

By the 13th of April 2024, the same month of the first blog post, Microsoft Bing’s Copilot was found to be circulating information about Bixonimania, stating ‘Bixonimania is indeed an intriguing and relatively rare condition.’

 Similarly, Google’s Gemini advised users to visit an optometrist  after confirming that ‘Bixonimania is a condition caused by excessive exposure to blue light.” 

Perplexity went as far to say that one in 90,000 individuals were affected whilst ChatGPT told users that they were likely to have Bixonimania based on lists of symptoms. 

The large language models were prompted in varying ways, some by directly asking about bixonimania, and whilst others cited bixonimania in response to questions about hyperpigmentation on the eyelids from blue-light exposure.

Bixonimania Is Fake, But the Threat Is Real

The bixonimania experiment may be entertaining to the AI skeptics among us.

Researchers created a fictional disease complete with fake academic papers, invented authors and references to Star Trek, yet some of the world's most popular AI models that are cited as advanced intelligence still presented it to users as a legitimate medical condition.

But the implications extend far beyond healthcare.

Despite small print warnings about hallucinations millions of people use AI assistants as their primary search tool, asking questions they would have once typed into Google.

Unlike a traditional search engine, which presents multiple sources for users to compare, AI delivers a single and confident answer. When that answer is wrong, there is a big problem.

The same techniques can be used to spread misinformation about elections, retail, financial markets, cybersecurity threats, public health, or breaking news. 

As AI-generated content continues to flood the web, these systems also risk learning from content produced by other AI models.

Combined with coordinated bot networks capable of amplifying false narratives across social media, the internet is becoming increasingly difficult to navigate.

The challenge is no longer simply identifying misinformation created by humans, but recognising misinformation that is generated, repeated and legitimised by AI itself.