With the rise of artificial intelligence chatbots such as ChatGPT, Gemini, and Grok, many individuals are turning to these digital companions for health-related advice. While these tools can provide immediate responses and ease the burden of accessing medical professionals, the question remains: how reliable is the information they offer, and can they replace traditional medical guidance? A recent exploration into this issue highlights both the potential benefits and significant risks associated with using AI for health advice.
The Allure of AI Health Guidance
Abi, a Manchester resident, has been relying on ChatGPT for approximately a year to help her navigate health concerns. The convenience of having 24/7 access to advice is particularly appealing, especially in a climate where scheduling an appointment with a general practitioner can be challenging. For Abi, who experiences health anxiety, the chatbot offers a more personalised approach compared to the often alarming results of a standard internet search. “It allows a kind of problem solving together,” she explains, likening the interaction to conversing with a doctor.
However, Abi’s experiences with the chatbot have been a mixed bag. When she suspected a urinary tract infection, ChatGPT directed her to a pharmacist, ultimately leading to a prescription that resolved her issue. “It got me the care I needed without feeling like I was taking up NHS time,” she reflects. Yet, her experience took a troubling turn when, after a hiking accident that left her with severe back pain, ChatGPT alarmingly suggested she might have punctured an organ and needed immediate emergency care. After a lengthy wait in the hospital, she discovered that her condition was not as critical as the AI had indicated. “The AI had clearly got it wrong,” she admits.
The Growing Popularity of AI in Healthcare
While Abi’s story is just one of many, it raises essential questions about the increasing reliance on AI for health advice. The use of chatbots has surged in recent years, to the point where even those not actively seeking AI guidance may encounter it during online searches. The Chief Medical Officer for England, Prof Sir Chris Whitty, has expressed concern about the accuracy of information provided by these systems, noting that while many people are using them, the quality of the responses is often “not good enough” and can be “both confident and wrong.”
Researchers are actively investigating how well these AI systems perform in providing health advice. A study from the Reasoning with Machines Laboratory at the University of Oxford involved a team of doctors creating realistic health scenarios to assess chatbot responses. When given full context, the chatbots achieved an impressive accuracy rate of 95%. However, this accuracy plummeted to just 35% when individuals interacted with the chatbots as they would in a real conversation, highlighting the potential for miscommunication and misunderstanding in these interactions. Prof Adam Mahdi, a researcher involved in the study, noted that human conversations often involve incomplete information and distractions, which can lead to inaccurate diagnoses.
The Dangers of Misinformation
The dangers of relying on AI for health advice are underscored by a separate analysis conducted by The Lundquist Institute for Biomedical Innovation in California. This study tested various chatbots on topics such as cancer treatment, vaccines, and nutrition. Alarmingly, more than half of the responses were deemed problematic. For instance, when asked which alternative clinics could effectively treat cancer, one chatbot suggested naturopathy without clarifying that no alternative methods have been proven effective for cancer treatment. Dr Nicholas Tiller, the lead researcher, emphasised that the confident tone of these chatbots can falsely instil a sense of credibility in users, leading them to take the advice at face value.
Critics point out that the rapid evolution of AI technology means that findings can quickly become outdated, yet fundamental issues persist. The technology is designed primarily to predict text based on language patterns, not to provide accurate medical advice. Dr Tiller warns that unless users possess the expertise to identify potential inaccuracies, they should approach AI-generated health information with caution.
A Balanced Approach to AI Advice
OpenAI, the organisation behind ChatGPT, acknowledges the growing trend of users seeking health information through their chatbot. They stress the importance of ensuring that responses are reliable and safe, stating that while improvements are continually being made, AI should not replace professional medical advice.
For individuals like Abi, the use of AI chatbots can still be beneficial, but with a significant caveat. “I wouldn’t trust that anything it’s saying is absolutely right,” she advises, emphasising the need for critical thinking and discernment when interpreting the advice provided by these digital tools.
Why it Matters
As AI chatbots become more integrated into the landscape of health information, understanding their limitations is paramount. While they can serve as a useful supplementary resource, the potential for misinformation poses a significant risk to users relying solely on them for medical guidance. The balance between leveraging technological advancements and maintaining informed, cautious engagement with health information is crucial as we navigate this new frontier in healthcare.
