More than half of ChatGPT’s active users now primarily use a language other than English, as shown in consumer usage data published by OpenAI. The most common languages other than English are Spanish, Portuguese, and Arabic.
The figures include individual ChatGPT packagessuch as Free, Go, Plus and Pro, but do not cover Codex, Enterprise or educational products. This means they do not reflect total usage in workplaces, technical areas or classrooms. Among this group of consumers, OpenAI says usage in languages other than English has increased alongside the platform’s overall growth since its 2022 launch.
What the data shows
OpenAI’s Signals program tracks how consumers use their plans over time. Measured against a July 2023 baseline, ChatGPT’s weekly active user count increased across all continents, with the fastest relative growth in Africa and Asia.
OpenAI reports that low HDI countries have seen the fastest relative growth during this period and notes that it continues to provide low-cost access through Free and Go plans.
Within this non-English business, it is the smaller languages that are growing the fastest. As of June 2026, Uzbek, Kazakh, and Burmese recorded the largest increases in active users among languages with more than 1 million users.
A separate analysis shows that after six months on ChatGPT, people send approximately 50% more messages daily and attempt twice as many unique tasks. This data, taken from a small 0.1% sample of accounts created between mid-October 2025 and early May 2026, provides a snapshot of recent user activity through the end of this period. Keep in mind, however, that it does not present the entire history of the platform.
OpenAI estimates that people with typically feminine names now lead among classifiable names, after reaching parity last year, by expanding the demographics inferred using name-gender matching rather than user data.
Why it matters
ChatGPT’s user base has shifted from a majority of English speakers to a more geographically diverse, less English-centric group than early adopters.
This correlates poorly with the behavior of the recovery layer. I covered a report earlier this year showing that ChatGPT Search often runs broadcast queries in the background in English, even when the original prompt is not in English, meaning that a question in Spanish or Arabic can still be pulled from English sources before getting an answer.
This is the second time in less than a year that OpenAI has released consumer usage data through policy research. When the company published a a previous large-scale review of consumer use with Harvard last Septemberthe results were a narrowing of the gender gap and faster growth in low-income countries. This version extends this image.
Looking to the future
Signals is an ongoing release, so it’s interesting to see how the non-English majority evolves from here.
The stakes are particularly high for companies that cater to an English-speaking audience. When most ChatGPT users use another language, an English-only audience is no longer the audience.
