Microsoft: Copilot Usage Varies by Device

Microsoft has shared new insights into how people use its Copilot AI assistant, revealing that behaviour changes noticeably depending on whether users are on a mobile device or a desktop computer.

The findings come from an analysis of 37.5 million Copilot conversations collected between January and September, offering a broad snapshot of how consumers interact with the tool across different contexts.

According to Microsoft, device choice plays a major role in shaping both the topics users explore and the type of help they seek from Copilot.

On mobile devices, Health and Fitness emerges as the most common topic throughout the entire day. This pattern remains consistent regardless of the time or month analysed.

Researchers note that mobile users are not simply looking for factual information. Instead, many are seeking guidance, reassurance, and practical advice, suggesting a more personal relationship with the assistant.

Desktop usage, however, follows a very different rhythm. While Technology remains a strong topic overall, conversations related to Work and Career dominate during standard office hours.

Between 8 a.m. and 5 p.m., work-related queries surpass technology discussions on desktop, reflecting how Copilot is used as a productivity tool during the working day.

The researchers describe three main modes of interaction: the workday mode focused on productivity, a constant personal companion mode linked to mobile use, and an introspective night mode that appears outside working hours.

During business hours, topics such as Education and Science also rise in prominence, particularly on desktop, aligning with professional learning and research activities.

Outside those hours, usage shifts noticeably. Late-night conversations tend to become more reflective, with topics like Religion and Philosophy appearing more frequently.

The analysis also highlights weekly and seasonal patterns. Programming-related discussions are more common during weekdays, while gaming conversations increase over the weekend.

Personal topics fluctuate throughout the year as well, with relationship-focused conversations spiking around occasions such as Valentine’s Day.

Microsoft explains that the study relied entirely on automated machine-learning classifiers to label conversation topics and intent, without any human review of individual chats.

The company also points out several limitations. The research is a preprint and has not yet been peer reviewed, and it focuses solely on consumer Copilot usage rather than enterprise traffic.

Overall, the findings suggest that Copilot is not used in a single, uniform way. Instead, how people interact with the assistant is shaped by their device, their schedule, and the role Copilot plays in their daily lives.