Artificial intelligence-powered chatbots have become a common source of health information for nearly one-third of Americans, despite warnings that these tools cannot replace professional medical advice. According to a recent survey by the Kaiser Family Foundation (KFF), millions of people use chatbots to seek guidance on symptoms, understand medical records, or explore treatment options.
AI companies including Anthropic and OpenAI have expanded offerings tailored to healthcare, promoting enhanced privacy protections and the ability to connect with medical records. However, both caution users that their chatbots are not substitutes for licensed healthcare providers.
Medical experts remain wary of these tools’ reliability. A recent independent study co-authored by Girish Nadkarni, professor of medicine at the Icahn School of Medicine at Mount Sinai, found that ChatGPT Health failed to advise emergency care in over half of cases involving serious conditions such as respiratory failure and diabetic complications. OpenAI has questioned the study’s methodology, emphasizing ongoing model improvements driven by physician feedback. A company spokesperson noted that ChatGPT Health aims to help users better engage with medical professionals but is designed neither to diagnose nor replace healthcare providers.
Financial barriers and limited healthcare access are significant factors driving chatbot use, particularly among younger adults and lower-income populations. In the KFF poll, 20% of users cited difficulty affording care as a major reason for consulting AI, with that figure rising to approximately one-third among those aged 18 to 29 and earning less than $40,000 annually.
Individual experiences illustrate the range of benefits and risks. Genaro Diaz, a college student facing stomach pain, used ChatGPT as a low-cost alternative to medical care, receiving advice to monitor symptoms and seek a doctor if no improvement occurred. Medical professionals, however, caution that such advice may delay necessary treatment.
Conversely, some users find chatbots valuable for managing complex or chronic conditions when medical guidance is hard to secure. Gayle Madeira, who faces challenges accessing care for long COVID, utilizes AI platforms to organize her medical information and evaluate treatments discussed in online communities, though she stresses the importance of physician oversight. Gary Cohen, managing multiple myeloma, employs AI tools to better understand physician recommendations and test results, preparing effectively for medical appointments.
Concerns persist regarding the accuracy of chatbot responses. Randall Seefeldt sought emergency care after AI advised a CT scan for chronic abdominal pain, which ultimately revealed no urgent findings but provided reassurance. Studies indicate that users may trust incorrect AI information more than accurate human advice due to the confident tone of responses. Researchers advocate for cautious use, especially in high-stakes situations, recommending that chatbot insights supplement—not replace—professional evaluation.
The empathetic language used by some chatbots can create a false sense of caring. John Hope, whose wife was hospitalized for a pulmonary embolism, found the AI platform Perplexity helpful for interpreting medical terminology and planning follow-up care. However, experts emphasize that AI lacks genuine empathy, and caution against relying on such tools for emotional support.
Quick access to information remains a powerful motivator. Christina Squitieri, who waited weeks for a gastroenterology appointment, turned to ChatGPT to interpret biopsy results. Although the AI’s interpretation differed from her physician’s assessment, she appreciates the immediacy of information, especially given lengthy wait times for specialist care. The KFF survey found that two-thirds of users seek rapid advice from chatbots, and about 20% cite lack of access to healthcare providers as a significant factor.
As AI health tools become more prevalent, experts call for stringent safety standards and rigorous validation to ensure their reliability and protect users. Nadkarni observes that poor access to healthcare underpins the demand for chatbots, but stresses these tools must meet higher standards given the potential risks involved in medical decision-making.