Should you really trust health advice from an AI chatbot?
Get the Health newsletter
Daily health & science — research, biotech, public health, the studies worth knowing. Free.
- ChatGPT correctly advised Abi to see a pharmacist for a suspected urinary tract infection, resulting in an antibiotic prescription without using NHS time.
- ChatGPT mistakenly told Abi she had punctured an organ after a back injury, prompting an unnecessary three‑hour A&E visit.
- Chris Whitty warned that AI health answers are 'not good enough' and often 'confident and wrong', highlighting systemic risk.
- Oxford’s Reasoning with Machines Lab reported chatbots achieved 95% accuracy when given full clinical scenarios, per researcher Adam Mahdi.
Why it matters: NHS patients can get faster triage from AI, but misdiagnoses risk costly emergency visits; 95% accuracy in tests masks real-world errors, potentially eroding trust in digital health.
