Abby has been relying on the AI chatbot ChatGPT for her health-related needs over the past year. Difficulty in securing appointments with general practitioners (GPs) and the need for quick responses have increased her interest in artificial intelligence chatbots. While AI has been found to pass certain medical exams with ease, can we place our trust in chatbots like ChatGPT, Gemini, or Grok? How does using these tools differ from traditional internet searches? And are experts’ concerns that their errors could put lives at risk justified?
Abby, from Manchester, has a keen interest in health topics. She finds it easier to ask questions directly to a chatbot rather than conducting internet searches, although it often points her toward the most alarming possibilities related to her symptoms. She explained, “It leads you toward problem-solving approaches.” When she suspected she had a urinary tract infection, ChatGPT advised her to consult a pharmacist. Following their advice, she received antibiotics. She described receiving essential guidance from the chatbot and said, “It’s very useful for people who are uncertain about when to see a doctor.”
However, in January, while going out, Abby slipped and fell. Her back hit a hard surface, and she felt a “deep” pressure across her back. She consulted the AI about this. ChatGPT suggested she may have sustained an injury and urgently advised her to visit emergency services. After spending three hours in emergency care, her pain gradually eased, and she was reassured she hadn’t suffered a serious injury. This instance showed that the AI had provided entirely incorrect information.
It is difficult to ascertain how many people, like Abby, use chatbots for health advice, but this technology’s popularity has become widespread. The Chief Medical Officer for England, Professor Sir Chris Whitty, has expressed serious concerns about the quality of advice provided by chatbots. He stated, “We have reached a critical point as more people use them.” He added that the information given by chatbots is not entirely reliable and that these chatbots often provide “confident but wrong” answers.
