OpenAI publishes first country-by-country ChatGPT usage data

For the first time, OpenAI released country-by-country data on how more than 1 billion ChatGPT users interact with the chatbot — and the numbers show a decisive shift from "asking" questions to "doing" tasks, especially at work.

Across the globe, ChatGPT users are more than twice as likely to use the tool to complete a task or produce output at work than they are outside it, according to the company's Economic Research team. The dataset, published on OpenAI Signals, covers 144 countries and pulls from messages sent through ChatGPT Free, Go, Plus, and Pro accounts — the individual-tier products that OpenAI says are "generally managed by individuals rather than organizations."

What does the new dataset actually measure?

The release is OpenAI's most granular public look at consumer behavior on the platform. It classifies messages into categories such as writing, coding, multimedia, practical guidance, and information-seeking. It tracks self-reported age, and it weights each country's share against population.

That weighting is what makes the per-capita leaderboard possible. OpenAI ranked countries by messages per capita in Q2 2026 and compared the standings against Q1 2026. The result is a moving map of where adoption is rising fastest.

The methodology document, released alongside the data, lays out how messages get classified and which countries are excluded for thin samples. Researchers can download the underlying CSVs from OpenAI's data hub. The Q2 release covers 126 countries with eligible multimedia estimates and 111 countries with complete age breakdowns.

Why does the work "doing" gap matter?

The dataset confirms a trend that has grown louder since OpenAI rolled out ChatGPT Work earlier this year: professional use is dominated by output, not queries. At work, users edit, code, draft, and run analyses through ChatGPT. Outside work, they ask. Seeking information and clarification remains the largest casual-use category.

The split matters because it changes what companies, regulators, and educators have to plan around. A tool used mostly to look things up has different downstream effects on labor markets, IP law, and classroom policy than a tool used to draft a contract or write a function.

For OpenAI, the work "doing" share is also a commercial signal. The company's pricing tiers have moved toward higher monthly fees for models that can complete long agent-style tasks. If most users are already executing work in ChatGPT, the path to monetizing those tasks becomes shorter.

Which countries are catching up to early adopters?

The headline change in Q2 2026: parts of the Southern Hemisphere closed ground on the early adopters. Peru, Uruguay, and Costa Rica climbed the furthest in the per-capita rankings. African and Oceanian regions posted faster growth rates than the global average.

Adoption in North America and Europe keeps rising, but the rate of increase is steeper in the Latin American and Pacific regions OpenAI tracks. The company frames this as the global adoption gap narrowing, a phrase its researchers used in a June update.

For policymakers, the question is whether infrastructure keeps up with usage. Peru and Uruguay both rank below the OECD average on fixed broadband penetration, according to ITU figures from 2025. Rising per-capita usage on weaker networks hints at mobile-first access patterns that regulators have not fully mapped.

What's driving the multimedia surge?

Since OpenAI shipped ChatGPT Images 2.0 in April 2026, multimedia messages have grown faster than any other category. Globally, multimedia now accounts for 7.8% of messages, up from a single-digit sliver earlier in the year.

The tilt is sharper in Latin America. In Brazil and Colombia, more than one in ten messages is now classified as multimedia. OpenAI's release includes multimedia estimates for 126 countries with eligible Q2 data.

Multimedia still trails writing, practical guidance, and information-seeking as a share of messages. But its consistent upswing through 2026 marks the first time a non-text category has commanded double-digit share in any major region OpenAI tracks.

The geographic skew raises a research question OpenAI does not answer: whether the multimedia tilt reflects a Latin American content preference, or whether ChatGPT Images 2.0 launched with features that landed differently in those markets.

Who is adopting ChatGPT now?

Older users. The share of messages from people aged 35 and older rose in nearly every country OpenAI tracks. Year over year, the over-35 cohort now sends a 5-percentage-point larger share of all messages than it did 12 months ago.

In France and Czechia, the over-35 share jumped more than 10 percentage points in a single year. Almost three-quarters of European countries posted above-average gains for this cohort. The pattern inverts in parts of Southeast Asia. Singapore showed a smaller increase, and six of the region's eight countries posted only marginal gains.

The data reflects only users who self-reported their age on the platform. That sample skews toward engaged users, OpenAI notes in its methodology document, which limits how broadly the trend can be generalized to the full billion-person user base.

Still, the direction is consistent. The early-adopter stereotype of a ChatGPT user — a 20-something developer in California — no longer matches the median sender.

What does OpenAI gain from publishing this?

The release puts OpenAI in line with public-interest research norms in a way consumer AI labs have rarely matched. Anthropic and Google DeepMind publish usage and capability reports, but rarely at country granularity for a billion-user product.

OpenAI positions Signals as a research hub for "policymakers and researchers around the world." The Q2 2026 release includes downloadable CSVs and a public data dictionary. That gives outside analysts the means to verify or contest OpenAI's framing.

The strategic motive is harder to miss. As governments from Brussels to Brasília draft AI rules, the company that can point to a country-by-country adoption map holds the louder microphone in those debates. The Signals launch also comes weeks after OpenAI published separate figures on business adoption, a pairing the company has used to argue its tools are now embedded across both consumer and enterprise markets.

What to watch next

The Q3 2026 update, due in the fall, will show whether the work "doing" share continues to climb, and whether the Latin American multimedia surge extends into other regions. OpenAI's researchers also flag agent-style usage — multi-step task execution that goes beyond a single prompt — as the next category they expect to break out of single-digit share.

For now, the dataset offers the clearest public picture yet of a chatbot that no longer behaves like a search box. It behaves like a colleague — one whose tasks, geography, and demographics are now visible down to the country.