OpenAI has released country-level data on how ChatGPT is used for the first time. At work, people are more than twice as likely to hand over an actual task, such as writing, coding, or analysis, than they are to simply ask for information. Regionally, Latin America, Africa, and Oceania stand out, and usage is spreading to people over 35[1].

At work, doing beats asking

The first point OpenAI highlights is the gap between work and non-work usage. In work settings, people are more than twice as likely to use ChatGPT to produce an output or perform a task, such as editing, coding, or running an analysis, than they are outside work[1].

Outside work, the picture is different. Asking, meaning seeking information or clarification, remains the largest category[1]. The same service is being used in two distinct modes: as a tool that gets work done, and as a place to look things up.

OpenAI notes that this shift has become more visible since the rollout of ChatGPT Work, with professional use growing[1].

Latin America and Oceania climbed the rankings

Country comparisons are based on a ranking of messages sent per capita. A smaller number means higher per-capita usage[2].

In the second quarter of 2026, growth in parts of Latin America, Oceania, and Africa outpaced the rest of the world. Peru, Uruguay, and Costa Rica gained the most positions in the global rankings[1]. Adoption keeps rising in North America and Europe, but the Southern Hemisphere is closing the gap on a per-capita basis.

This did not begin overnight. The first-quarter update already showed countries outside the established markets, including the Dominican Republic, Haiti, Mexico, and Tanzania, moving up the list[3]. Japan also rose 8 places from 43rd to 35th in that period, placing it among the 10 fastest-rising countries[3].

Multimedia use grew to 7.8 percent

The fastest-growing use case is multimedia. Since ChatGPT Images 2.0 was released in April 2026, multimedia has grown to 7.8 percent of messages worldwide[1].

That is still behind the leading categories of practical guidance, writing, and information-seeking. Even so, the share has climbed consistently through the year, and in Brazil and Colombia more than one in ten messages is classified as multimedia[1]. Generating, analyzing, and retrieving images is becoming as routine as text.

Users over 35 are gaining share

Age patterns are shifting as well. The share of messages sent by people over 35 rose in nearly every country, and these users now account for a share 5 percent higher than 12 months earlier[1].

Europe stands out. In France and Czechia, the share of messages from users aged 35 and older increased by more than 10 percentage points in a single year. Almost three-quarters of European countries saw larger-than-average increases, while some Southeast Asian markets such as Singapore moved the other way[1]. In that region, 6 of 8 countries saw the share rise, but only by a small margin[1].

How to read the numbers

The data is published on OpenAI Signals, the platform run by the company's Economic Research team, and is free to browse or download[1]. It is based on a sample of messages sent between July 2024 and June 2026[2].

The scope matters. The dataset covers only messages from the individual ChatGPT plans, namely Free, Go, Plus, and Pro. Accounts managed by companies or educational institutions, along with Codex usage, are excluded[1][2]. Business and technical use is therefore likely understated, and the finding that work usage skews toward doing should be read as a comparison within consumer plans.

Summary

The first country-level release shows ChatGPT moving from a tool for finding answers to one for getting work done, with adoption broadening beyond early markets and older age groups. The 7.8 percent multimedia share is a clear case of a new capability changing how the product is used. Because the analysis is limited to individual plans, enterprise behavior is better checked against the separate Enterprise dataset.

[1] https://openai.com/index/how-the-world-is-putting-chatgpt-to-work

[2] https://openai.com/signals/data/

[3] https://openai.com/signals/research/2026q1-update/