Should You Trust AI With Your Health?

The appeal of AI is clear: it is free, available at any hour, and feels personal. Abi, a young woman from Manchester quoted by the BBC (British Broadcasting Corporation), said her chatbot once correctly suggested she see a pharmacist for a urinary tract infection — but later wrongly told her a back injury had punctured an organ, sending her to a hospital emergency room she did not need. Researchers from the University of Oxford found that chatbots reach 95% diagnostic accuracy when given complete medical scenarios, yet that accuracy collapsed to just 35% in real user conversations, where people tend to leave out important details. England's chief medical officer, Professor Sir Chris Whitty, warned that AI answers were often "both confident and wrong", a view echoed by separate research from the Lundquist Institute, which classed more than half of chatbot responses as problematic.

Doctors and health authorities are urging caution. Professor Victoria Tzortziou Brown, president of the Royal College of General Practitioners (RCGP), described the trend as "highly concerning", noting that AI cannot examine patients, read subtle physical signs, or make safe clinical judgments based on a full medical history. She encouraged the public to use trusted sources such as the NHS website and to contact a qualified professional when symptoms persist. OpenAI, the company behind ChatGPT, said its tool was designed to provide information and education rather than replace professional advice. The study's lead author, Professor Graham Lord, warned that growing chatbot use is creating "an unregulated AI healthcare system alongside the NHS" — one that may demand new rules to keep patients safe.