Nature Study Reveals Microsoft Copilot Health Query Usage Patterns from 500,000 Conversations

A Nature Health study analyzed 617,827 health-related conversations with Microsoft Copilot, revealing that Health Information and Education queries dominate at 40.8% of interactions, followed by significant platform-specific differences in usage patterns. Mobile users ask more Symptom Questions (15.9% vs 6.9% on desktop) and Emotional Well-being queries (5.1% vs 3.0%), while desktop users focus on Research and Academic Support (16.9% vs 5.3%) and Medical Paperwork (15.7% vs 2.7%). The study found temporal variations with symptom and emotional well-being queries rising from morning to nighttime, and revealed that a substantial fraction of personal health queries serve caregiving rather than individual needs. Microsoft emphasizes that Copilot is not intended to replace professional medical advice.

Health Query Patterns and Intent Categories in AI Conversations

• ▪The Health Information and Education category in Microsoft Copilot captures non-personalized queries including how medications work, what causes conditions, and general nutrition information.
• ▪The Microsoft Copilot health intent taxonomy comprises 12 primary categories with fine-grained topic clusters within each category.
• ▪Health Information and Education accounts for 40.8% of Microsoft Copilot health conversations with a 95% confidence interval of 40.7-40.9.
• ▪99.6% of the Microsoft Copilot health conversations analyzed had a valid timestamp and were included in temporal analyses.
• ▪A Nature study analyzed 617,827 health-related conversations from Microsoft Copilot after applying classification filters and excluding two categories.
• ▪99.1% of the Microsoft Copilot health conversations analyzed had a known platform value and were included in mobile-versus-desktop analysis.

Device and Temporal Variations in Health-Related AI Usage

• ▪Symptom Questions and Health Concerns account for 15.9% of Microsoft Copilot conversations on mobile versus 6.9% on desktop.
• ▪Symptom Questions and Health Concerns in Microsoft Copilot rise from 10.6% in the morning to 13.4% at nighttime with chi-square test results of χ²(23, N=613,026)=1,445.8 and P<0.001.
• ▪Emotional Well-being queries in Microsoft Copilot increase from 3.3% in the morning to 5.2% at nighttime with chi-square test results of χ²(23, N=613,026)=903.3 and P<0.001.
• ▪Emotional Well-being queries account for 5.1% of Microsoft Copilot conversations on mobile versus 3.0% on desktop.
• ▪Research and Academic Support queries account for 16.9% of Microsoft Copilot conversations on desktop versus 5.3% on mobile.
• ▪Desktop Microsoft Copilot usage is more predominant during the day while mobile usage is more predominant at night.
• ▪Intent distributions in Microsoft Copilot health conversations differ significantly between mobile and desktop platforms with chi-square test results of χ²(8, N=612,330)=73,981.6 and P<0.001.
• ▪Medical Paperwork queries in Microsoft Copilot peak during normal working hours.

Personal Health Concerns and Caregiving Through Conversational AI

• ▪A substantial fraction of Microsoft Copilot personal health queries serve caregiving rather than individual needs.
• ▪The Nature study analyzed a subsample of 2,165 Microsoft Copilot conversations drawn from the main dataset and annotated for the person the query concerns.
• ▪Microsoft Copilot is not intended to replace professional medical advice.
• ▪The Nature study subsample examining who health queries are about comprises three personal health intents: Symptom Questions and Health Concerns, Condition Management, and Emotional Well-being.

Perspective of Healthcare professionals

• ▪The prevalence of Symptom Questions and Health Concerns queries in Microsoft Copilot, especially at nighttime, may indicate users are turning to AI before or instead of consulting healthcare professionals.

Perspective of Digital health researchers

• ▪The significant platform differences in Microsoft Copilot health query types reveal that mobile and desktop contexts shape how users seek health information.

1 source