AI medical advice is filling a gap left by overwhelmed healthcare systems. The problem is that gap exists for a reason—doctors have training, liability, and accountability. ChatGPT has none of those.
New research from King's College London reveals that 15% of people have used AI chatbots for health advice instead of consulting a GP or NHS service. Another 10% used AI for mental health support instead of seeing a trained professional. And among those who did consult AI, 20% said it discouraged them from seeking professional healthcare.
Read that again. One in five people who asked an AI about their symptoms decided not to see a doctor because the AI told them they were fine. What happens when the AI is wrong?
It's wrong a lot, by the way. Recent studies indicate that AI chatbots misdiagnose in up to 80% of early medical cases. That's not a typo. Eighty percent. If a human doctor had that failure rate, they'd lose their license. But AI doesn't have a license to lose.
This is both a healthcare access story and an AI safety story. The reason people are turning to AI isn't because it's better than doctors—it's because doctors are inaccessible. In the UK, GP wait times can stretch weeks. In the US, many people are uninsured or underinsured. If your choice is "wait three weeks for an appointment" or "ask ChatGPT right now," it's understandable why people choose the latter.
But understanding why people do it doesn't make it safe. AI chatbots are language models trained on text. They generate plausible-sounding answers based on patterns in their training data. Sometimes those answers are accurate. Sometimes they're confidently wrong. The AI doesn't know which is which, and neither does the user.
Here's an example: someone asks , The correct answer is: ". Those are symptoms of a potential heart attack.It could be anxiety or indigestion. Try relaxing and drinking water. If symptoms persist, consider seeing a doctor."
