Patients Increasingly Turn to AI Chatbots for Medical Information, Despite Accuracy Concerns
Washington, D.C. – A growing number of patients are utilizing artificial intelligence (AI) chatbots like ChatGPT, claude, adn Gemini to interpret medical test results and records, notably when facing delays in reaching their physicians, according to emerging trends and recent data. patients are reportedly uploading sensitive health information – including lab results and imaging reports – to these platforms seeking explanations while awaiting doctor callbacks or appointments.
A recent KFF August 2024 Tracking Poll reveals that approximately 17% of adults report using AI chatbots at least monthly to seek health information and advice. This figure rises to 25% among adults under the age of 30. However, the poll also highlights significant skepticism regarding the accuracy of this information, with 63% of adults expressing “not too confident” or “not at all confident” levels of trust in health information sourced from AI chatbots. Conversely, 36% indicated some level of confidence, with 5% being “very” confident and 31% “somewhat” confident.
While AI chatbots can potentially aid patients in understanding their health data and reducing anxiety, medical professionals and researchers are cautioning against reliance on these tools due to inherent risks. A key concern is the phenomenon of AI “hallucinations,” where chatbots generate plausible but factually incorrect information, often presented with the same authoritative tone as accurate data. this makes it difficult for individuals without medical training – and even for medical professionals – to identify errors. A study published in BMC Medical Education in March 2025 found that general practice trainees achieved a mean accuracy of only 55% in detecting AI-generated medical hallucinations.
Recent research suggests that the accuracy of AI-generated responses can be improved through refined prompting strategies. A study in JAMIA Open from April 2025 demonstrated that instructing a chatbot to adopt the persona of a clinician enhanced accuracy. Further, an August 2025 Communications Medicine study showed that incorporating safeguards into prompts, such as requesting the AI to rely solely on clinically validated information, reduced the occurrence of hallucinations. Researchers suggest that educating users on effective prompting techniques could improve the utility of these tools, but emphasize that these strategies do not eliminate errors entirely. Experts have recommended that AI chatbots should be used as supplementary tools, rather than primary sources of health information, as highlighted in recent recommendations published in PMC.
