How People Use Copilot for Health
In a recent analysis of over 500,000 de-identified health-related conversations with Microsoft Copilot from January 2026, researchers sought to characterize what users ask conversational AI about health. The study, detailed in arXiv:2604.15331v1, developed a hierarchical intent taxonomy of 12 primary categories using privacy-preserving large language model (LLM)-based classification, which was validated against expert human annotation. Additionally, the researchers applied LLM-driven topic clustering to identify prevalent themes within each intent.
Key Findings
The study revealed five major findings about health-related queries made to Copilot, which highlight the diverse ways in which people utilize AI for health information.
- Personal Symptom Assessment: Nearly one in five conversations involved personal symptom assessment or discussions about specific health conditions. This indicates a significant interest in personal health management and suggests that the actual volume of personal health-related inquiries may be even higher than recorded.
- Caregiving Queries: One in seven personal health queries pertained to individuals other than the user, such as children or partners. This indicates that conversational AI is not only a tool for personal health inquiries but also serves as a valuable resource for caregivers seeking information about the health of loved ones.
- Increased Evening Usage: There was a marked increase in personal health queries, particularly about symptoms and emotional health, during evening and nighttime hours. This trend suggests that users may turn to AI when traditional healthcare options are less accessible, highlighting a potential gap in after-hours healthcare support.
- Device Usage Variance: The study found a significant divergence in usage patterns based on the device. Mobile devices were primarily used for personal health concerns, while desktop usage was more focused on professional and academic inquiries, indicating different contexts for health-related questions.
- Healthcare Navigation: A considerable portion of queries was dedicated to navigating healthcare systems, including finding healthcare providers and understanding insurance coverage. This underscores existing challenges in the healthcare delivery process and the role of AI in addressing these barriers.
Implications for AI Development
The patterns observed in this study have direct implications for the design and development of health AI platforms. Understanding user intent and behavior can inform the creation of more targeted, user-friendly applications that prioritize safety and accessibility. As conversational AI continues to evolve, these findings emphasize the need for responsible innovation that addresses the complexities of health-related inquiries and the varying needs of users.
In conclusion, the insights gained from this analysis not only enhance our understanding of how individuals interact with AI for health information but also point to the potential for these technologies to improve healthcare delivery and support both personal and caregiving needs.