A survey found that office workers use ChatGPT differently by country and that workplace uses are becoming more diverse. In Japan, use for creative thinking increased, while decision-making and problem-solving remained the main focus in the United States.

On Oct. 27 local time, Japanese media outlet GIGAZINE reported that OpenAI recently released a report analysing workplace use of ChatGPT in about 100 countries. The survey covered work-related messages exchanged on the consumer version of ChatGPT from the first quarter of 2025 to the third quarter of 2026.

OpenAI classified work into 165 activities based on the U.S. Occupational Information Network, O*NET, and analysed how usage differed by country and how it changed over time.

In South Korea, decision-making and problem-solving accounted for the largest share in the first quarter of 2025. They were followed by interaction with computers, information gathering and creative thinking, but by the third quarter this year the order shifted to decision-making and problem-solving, creative thinking, and documentation and information recording.

In Japan, there was a clear shift in the purpose of workplace use. In the first quarter of 2025, decision-making and problem-solving had the largest share, followed by interaction with computers, but by the third quarter this year the share of decision-making and problem-solving fell to 15.9 percent and creative thinking became the most used area. In the United States, decision-making and problem-solving remained the largest share over the same period. That showed that even the same AI service can be used differently for work depending on the country.

Globally, usage areas became more diverse. In 74 countries comparable across the full survey period, the average share of the top 10 work activities fell by 4.9 percentage points to 51.1 percent in the third quarter of 2026 from 56.0 percent in the first quarter of 2025.

The data also found a gap between AI use and actual labour market structure. Planning and decision-making accounted for 39.6 percent of U.S. work-related ChatGPT messages, but made up only 9.3 percent of all work activities reflecting the U.S. employment structure.

By contrast, other work activities including driving vehicles and transporting goods accounted for 43.1 percent of all U.S. work activities, but represented only 2.4 percent of ChatGPT use.

The data are part of OpenAI Signals, which releases information on AI usage. It excluded usage of the enterprise version and Codex, and it released only automatically classified statistics after removing identifying information to protect personal data.