As generative artificial intelligence tools like DeepSeek and Doubao become widely accessible in China, patients are increasingly arriving at clinical encounters already armed with AI-generated explanations of their symptoms, treatments, and care plans. A new qualitative study published in Frontiers in Public Health suggests this shift is reshaping nurse–patient communication in ways that both challenge and potentially enrich clinical interactions.

The study, conducted by researchers at Suining Central Hospital and Tongji University in Shanghai, interviewed 18 clinical nurses from three tertiary hospitals in Sichuan Province and Shanghai between May and August 2026. All participants had at least one year of clinical experience and had encountered patients or family members who explicitly mentioned, showed, or referred to AI-generated health information during a nurse–patient interaction in the previous 12 months.

According to the study, two main themes emerged from the nurses' accounts. The first, "changes in nurse–patient communication in the context of AI," captured how patients used AI-generated information to seek confirmation from nurses or to question clinical decisions. Nurses reported that this often required additional explanatory work — explaining why a general AI recommendation might not apply to a patient's specific condition, medication regimen, or treatment stage. At the same time, some nurses noted that patients who brought in AI information were more engaged and asked more questions, which could open up new communication opportunities.

The second theme, "nurses' responses to AI-generated health information," described how nurses adapted their communication strategies. Rather than dismissing the AI information outright, nurses reported trying to identify the patient's actual concerns behind the query, explain discrepancies between AI-generated information and clinical arrangements in light of the patient's specific situation, and seek professional support — such as consulting a physician or pharmacist — when the AI-generated content touched on areas outside the nurse's expertise.

For university students studying nursing, medicine, or public health, the findings carry direct relevance. As future healthcare professionals, they will likely encounter patients who have consulted AI tools before arriving at a clinic or hospital. The study suggests that effective communication in this new context requires not only clinical knowledge but also the ability to understand how patients have interpreted AI-generated information and to address the expectations that may have developed as a result.

The researchers noted that even when an AI-generated statement is broadly accurate, its relevance depends on the patient's individual clinical circumstances. "The factual accuracy of general health information does not necessarily mean that it is appropriate for a particular patient," the study authors wrote, pointing out that a patient's condition, medication use, treatment stage, and care plan may differ from the assumptions underlying an AI-generated response.

The study used a qualitative descriptive design with semi-structured interviews and inductive reflexive thematic analysis. The researchers employed purposive sampling combined with a maximum variation strategy to ensure diversity across clinical departments. Participants were recruited from a range of wards, including surgical, medical, and specialist units.

While the study provides detailed accounts of nurses' experiences, the researchers acknowledged limitations. The findings are based on nurses' retrospective accounts rather than direct observation of patient–nurse interactions, and the sample was drawn from three tertiary hospitals in two Chinese regions, which may limit applicability to other clinical settings or countries. The study's authors called for further research to explore how these communication patterns play out in different healthcare contexts and to develop training programs that help nurses respond effectively when patients bring AI-generated information into clinical conversations.

The study adds to a growing body of research examining the intersection of generative AI and healthcare communication. A 2024 survey of 2,406 respondents in the United States cited in the paper found that 21.5% had used ChatGPT to obtain health information, and some users reported that the generated information influenced their consideration of treatment adjustments or prompted them to seek further confirmation from physicians. The World Health Organization has also highlighted potential risks associated with using AI-generated information in health-related decision-making.

For nursing students and early-career nurses, the study underscores the importance of developing communication skills that go beyond simply explaining medical information. As one nurse quoted in the study described, the task now includes understanding what patients already know, how they came to know it, and how to bridge the gap between general AI-generated advice and individualized clinical care.