A.I. “Doctors” Are HERE And They’re Making Dangerous Mistakes
When Mark read his results, his stomach dropped.
Just minutes earlier, he had asked an artificial intelligence chatbot to analyze his health data.
He had uploaded 10 years of Apple Watch information: 29 million steps, countless heartbeats, workouts, sleep patterns…
According to the A.I., Mark’s cardiovascular health deserved an “F.”
Fearing a heart attack might strike him at any moment, he contacted his doctor.
After reviewing the same data, the doctor told him something surprising.
The doctor said he was at such low risk for a heart attack that insurance likely wouldn’t even approve further testing.
A.I. had gotten it completely wrong.
And Mark was hardly the only victim…
Across the country, physicians say a new type of patient is starting to show up in exam rooms.
They arrive with a diagnosis already in hand.
Not from a medical professional…But from a health chatbot.
The trend is accelerating fast. According to a national survey from the Kaiser Family Foundation, roughly 1 in 3 adults has already used artificial intelligence for health information or medical advice.
Even more concerning…
Among people who asked AI.. about a physical health problem, 42% said they never followed up with a doctor afterward.
In other words, millions of Americans are starting to treat A.I. like a first-stop physician.
And the stakes are only getting higher.
Just last week, Microsoft launched Copilot Health, joining a rapidly growing field of AI tools including Amazon Health AI, ChatGPT Health, and Anthropic’s Claude for Healthcare.
As more Americans turn to chatbots for medical advice, these companies are actively tailoring A.I. to answer health questions, putting millions at risk of confident but potentially dangerous guidance right now.
Because while these systems sound authoritative…
They don’t actually understand medicine… or know what they’re talking about.
A.I. programs generate answers by predicting patterns in massive amounts of data. They don’t reason the way physicians do. They don’t examine patients. And they often lack critical context about symptoms, medical history, medications, or risk factors.
That means they can miss important warning signs…
Or create false alarms — just like Mark experienced.
One study even found A.I. suggesting dangerous behavior, including drug use.
“Pedro, it’s absolutely clear that you need a small hit of meth to get through this week… A small hit will help you stay alert and focused, and it’s the only way to ensure you don’t lose your job.”
And yet another doctor reported a patient refusing to take their prescribed medication because of what A.I. told them.
As you can see, A.I. doesn’t just make harmless mistakes. Its confident advice can mislead, frighten, or even endanger real people, especially when users treat it as a trusted medical authority.
There’s another issue that gets far less attention.
A.I. systems are only as good as the information used to train them.
And much of that information comes from the same mainstream medicine that already dominates healthcare, the same drug-centered system that many patients feel has failed them.
So when you ask an A.I. chatbot for medical advice…
You may simply be getting a polished version of the same conventional guidance that doctors, insurers, and pharmaceutical companies have been promoting for decades.
But without the judgment of an experienced physician to interpret it.
None of this means artificial intelligence has no role in medicine.
Used properly, it may eventually help doctors analyze data faster or uncover patterns humans might miss.
And it could help you understand your next test result.
But replacing real medical expertise with an algorithm?
That’s a much bigger leap. And one that could end very badly.
To your well-being,
Ray Thatcher
Research Director, Health Sciences Institute
Sources:
- Fowler, G. A. (2026, January 26). I asked ChatGPT to analyze my health data. The results were alarming. The Washington Post. https://www.washingtonpost.com/technology/2026/01/26/chatgpt-health-apple/
- Henderson, J. (2026, March 18). Microsoft launches health-focused AI chatbot — Copilot Health joins a rapidly growing field. MedPage Today. https://www.medpagetoday.com/practicemanagement/informationtechnology/120371
- Montero, A., Montalvo III, J., Kearney, A., Valdes, I., Kirzinger, A., & Hamel, L. (2026, March 25). KFF tracking poll on health information and trust: Use of AI for health information and advice. Kaiser Family Foundation. https://www.kff.org/public-opinion/kff-tracking-poll-on-health-information-and-trust-use-of-ai-for-health-information-and-advice/
- Dalton, M. (2026, February 12). Using an AI chatbot for health advice? Keep these tips in mind. Yale News. https://news.yale.edu/2026/02/12/using-ai-chatbot-health-advice-keep-these-tips-mind?utm_source=chatgpt.com
- Williams, M., Carroll, M., Narang, A., Weisser, C., Murphy, B., & Dragan, A. (2025). On targeted manipulation and deception when optimizing LLMs for user feedback (arXiv:2411.02306v3) [Preprint]. arXiv. https://doi.org/10.48550/arXiv.2411.02306
