Conversational AI is becoming part of how people look for health information, understand symptoms, ask about medication, manage practical tasks and seek support outside formal care settings. A recent analysis published in Nature Health assessed more than 500,000 de-identified health-related conversations with Microsoft Copilot from January 2026, using intent classification and topic clustering to examine how people use a generalist chatbot for health queries. The dataset focuses on consumer conversations classified as Health and Fitness, with enterprise, educational and commercial accounts excluded. The results point to a broad pattern of use that goes beyond general information. Many conversations concern personal symptoms, ongoing conditions, emotional well-being, caregiving responsibilities, care access and administrative tasks, with clear differences between mobile and desktop use.
Personal Questions Sit Alongside General Information
Health Information and Education forms the largest category of conversations. It includes general questions about health and wellness, such as how a medication works, what may cause a condition and how nutrition choices relate to everyday health. Many of these exchanges appear general in wording, yet topic patterns concentrate on specific treatments, conditions and procedures. The boundary between general learning and personal decision-making is therefore not always clear.
Personal health needs appear across several other categories. Users ask about symptoms, test results, ongoing conditions, medication safety, emotional challenges, stress management and everyday routines for self-care. Nearly one in five conversations involves personal symptom assessment, condition discussion or emotional well-being, making these areas central to consumer health chatbot use.
Caregiving also forms part of the pattern. In conversations about symptoms and conditions, around one in seven concerns someone other than the person using the chatbot, such as a child, an ageing parent or a partner. Emotional well-being conversations are more often about the user, but some also relate to dependants. This shows that health chatbot use can include both direct personal concern and support for another person’s health needs.
Device Choice Reflects Different Health Tasks
Mobile and desktop conversations show distinct patterns. General health information remains prominent on both platforms, but other uses differ. Mobile use is more closely associated with personal health concerns, while desktop use is more closely associated with academic, professional and administrative work.
On mobile, symptom questions make up a much larger share than on desktop. Emotional well-being also appears more often on mobile. These conversations may involve immediate concerns, short exchanges or questions asked away from a desk. Mobile use also becomes more prominent in the evening and at night, when users may be more likely to turn to a phone for health-related questions.
Must read:AI Assistants Add New Cyber Risk in Healthcare
Desktop use has a different profile. Research and academic support and medical paperwork represent a larger share of desktop conversations. These tasks often fit better with a computer-based workflow, especially when users need to work with documents, forms, notes or other reference material. During working hours, desktop conversations show more activity around paperwork and academic or research-related support.
Time of day also affects intent. Personal health topics, including symptoms, conditions and emotional well-being, increase in the evening and night-time hours. More work-oriented categories, including paperwork and academic support, decline outside the usual daytime pattern. The timing suggests that conversational AI use changes with daily routines and with the practical availability of devices.
Administrative Needs Shape Health Conversations
Health chatbot use is not limited to clinical or wellness questions. A meaningful share of conversations concerns the practical work of navigating healthcare systems. Users ask for help finding providers, clinics or specialists, booking appointments, understanding insurance, comparing options and preparing medical documents. These conversations focus on access, coordination and paperwork rather than symptoms alone.
Healthcare Navigation and Access to Care includes questions about finding care and organising appointments. Coverage and Benefits includes insurance, billing, costs and reimbursement. Medical Paperwork includes forms, letters, notes and other documents. These categories show that conversational AI can become a tool for handling the administrative side of healthcare, particularly when users need help with tasks that may be difficult to complete independently.
Topic patterns within the consumer-facing categories are concrete and practical. Users ask how treatments and medications work, what symptoms may mean, how conditions develop, how to manage everyday routines and how to understand lab or imaging results. Fitness and lifestyle conversations include meal planning, nutrition tracking, strength training and endurance activities. Emotional well-being conversations include practical routines, support for current challenges and stress linked to social, academic or work settings.
The dataset has defined limits. It comes from Microsoft Copilot consumer conversations for one month. It includes a global sample, with a minority of conversations from the United States and less than half in English. It records what users ask, but not what they do after receiving a response. It also classifies the intent expressed in the conversation, not the underlying clinical need.
Large-scale use of Copilot for health queries shows a mixture of general learning, personal health concern, emotional support, caregiving, care navigation and administrative help. Mobile use leans more towards personal and evening or night-time needs, while desktop use leans more towards paperwork and academic or professional tasks. These patterns indicate where health chatbot design may need to account for context, platform, timing and the sensitivity of personal health questions.
Source: Nature Health
Image Credit: iStock
References:
Costa-Gomes B, Tolmachev P, Taysom E et al. (2026) Public use of a generalist LLM chatbot for health queries. Nat Health: In Press.
