How people use Copilot for Health

A comprehensive analysis of over 500,000 de-identified health conversations with Microsoft Copilot reveals that conversational AI serves dual roles in healthcare—personal symptom assessment and caregiver support—with usage patterns heavily influenced by device type and time of day. The research demonstrates that 20% of queries involve personal health concerns, while 14% address health questions about others, underscoring AI's expanding role in informal healthcare delivery and system navigation.

Microsoft's analysis of 500,000+ Copilot health conversations provides empirical evidence of how conversational AI is reshaping healthcare information-seeking behavior. The study applies a validated 12-category taxonomy to characterize user intents, revealing that despite general information comprising 40% of queries, most focus on specific treatments and conditions—indicating that personal health concerns represent a substantial floor rather than the stated 20%. This distinction matters because it demonstrates users are treating conversational AI as a quasi-clinical advisor, not merely an encyclopedia.

The finding that 14% of personal health queries concern third parties—children, parents, partners—reframes the value proposition beyond individual consumers to family caregiving networks. This reflects broader demographic shifts toward aging populations and caregiving burden. The temporal concentration of symptom and emotional health queries during evening and nighttime hours directly addresses a healthcare access gap, as traditional clinical services operate limited hours. Mobile-desktop divergence reveals distinct use cases: mobile users pursue personal health information while desktop users engage in professional and academic work, suggesting platform-specific opportunities for targeted features and interventions.

The substantial focus on healthcare system navigation—insurance, provider discovery, billing—highlights critical friction points in existing healthcare delivery that conversational AI can potentially alleviate. For the AI industry, this validates commercial opportunities in health-tech applications. For healthcare providers and policymakers, it signals both the potential benefits and regulatory risks of unmediated AI health advice. The responsible development emphasized by researchers points toward hybrid models combining conversational AI with clinical oversight, rather than standalone deployment.

• →Nearly 20% of Copilot health conversations involve personal symptom assessment, with 14% addressing health concerns about family members or dependents.
• →Evening and nighttime usage spikes for symptom and emotional health queries, addressing critical gaps when traditional healthcare access is most limited.
• →Mobile devices concentrate on personal health queries while desktop usage focuses on professional and academic health-related work.
• →Significant query volume targets healthcare system navigation—insurance, provider finding, billing—revealing substantial friction in existing healthcare delivery.
• →Validated hierarchical intent taxonomy and privacy-preserving LLM classification methodology provides replicable framework for health AI research and safety auditing.