ChatGPT's Non-English Majority: What OpenAI's Signals Data Reveals About Global AI Adoption

OpenAI's Signals initiative, a public dataset launched in 2025, provides an unprecedented window into how people actually use ChatGPT. Drawing on a privacy-preserving sample of millions of consumer conversations, the data paints a clear picture: ChatGPT has gone global, and the English-speaking world is no longer its center of gravity.

Here is what the latest Signals data and related reporting reveal about the languages, regions, and challenges shaping ChatGPT's global footprint.

Non-English Usage Now Exceeds English on ChatGPT

Users who predominantly communicate in a language other than English now represent over half of all active ChatGPT users . The three leading non-English languages on the platform are Spanish, Portuguese, and Arabic .

The Signals dataset, which samples 300,000 consumer messages per month from July 2024 onward, categorizes usage by region, language, purpose, and user group . An OpenAI blog post confirms that "users predominantly using a language other than English now represent over half of active users" .

The Fastest-Growing Languages: Azerbaijani and Kazakh

According to OpenAI CFO Sarah Friar, speaking on the All-In Podcast at the Liquidity Summit in Hong Kong in June 2026, Azerbaijani and Kazakh are the fastest-growing languages among consumers on ChatGPT . Friar described the trend as "kind of incredible" .

OpenAI's own blog further identifies Uzbek and Burmese among the languages with the largest percentage increase in their share of active users since July 2023 . The fastest-growing region overall is the African continent .

Geographic Regions Leading Adoption

OpenAI Signals data and external analytics reveal the following adoption hotspots:

• The United States leads with approximately 15.1% of ChatGPT traffic .
• India is the second-largest market and the fastest-growing major market .
• Brazil ranks third, accounting for 5.3% of traffic .
• The African continent is the region with the fastest overall user growth .
• ChatGPT is available in 161+ countries and supports 95+ languages .

India's significance goes beyond raw user numbers. OpenAI announced plans to open new offices in Bengaluru and Mumbai in 2026, expanding beyond its existing base in New Delhi. Sam Altman confirmed this at the AI Impact Summit in February 2026 . India is now OpenAI's second-largest market, with over 10 crore (100 million) weekly ChatGPT users, including the world's largest student user base .

OpenAI's Expansion Into Non-English Markets

OpenAI's global push goes beyond language support into infrastructure and education.

Education for Countries: Launched at the World Economic Forum in Davos in January 2026, the program partners with eight nations: Estonia, Greece, Italy, Slovakia, Trinidad and Tobago, Kazakhstan, the UAE, and Jordan . The initiative focuses on bringing AI into education systems to personalize learning, reduce administrative burden, and prepare students for an AI-driven workforce .

OpenAI Academy: A free online learning platform that reached over 2 million learners .

OpenAI Certifications: Launched to provide credentials for AI fluency, from basics to advanced prompt engineering and AI-custom jobs .

ChatGPT Go: The low-cost subscription plan expanded to 170+ countries after its initial launch in India in August 2025, making it OpenAI's fastest-growing plan .

ChatGPT's Huge User Base: 900 Million Weekly Active Users

ChatGPT reached 900 million weekly active users in February 2026 — more than doubling the 400 million weekly user base from February 2025 . The platform processes approximately 2.5 billion prompts per day and 18 billion messages per week .

Persistent Technical Challenges for Non-English Users

Despite the progress, OpenAI's Signals data and independent research highlight several technical gaps that remain for non-English speakers:

• Transcription and speech accuracy gaps: While recent updates improved dictation accuracy for Japanese, Korean, Chinese, Urdu, Vietnamese, accented English, and long-form Spanish, French, Italian, and Portuguese, overall transcription accuracy — especially across mixed-language conversations — is still under active improvement .

• Inconsistent language output: Users report that non-English language output can be inconsistent, depending on device settings, app language, and context. OpenAI's community forums acknowledge this but note there is no formal policy fix .

• Lower accuracy for low-proficiency English users: Research from the MIT Center for Constructive Communication found that leading AI models (including ChatGPT) perform worse for users with lower English proficiency, less formal education, and non-US origins — meaning the same query can yield less accurate results depending on the user's background .

• Non-Latin script quality gaps: Independent research has documented that AI systems excel at translating various languages into English but struggle to convert English into languages using non-Latin scripts, such as Korean .

• Language detection instability: Model version transitions for GPT-4o-transcribe introduced instability in language detection, sometimes misidentifying the input language .

These challenges are not new, and OpenAI has acknowledged biases in its systems . The gap between language support in the interface — which covers 50+ languages officially, with practical support for over 95 — and true parity in accuracy and reliability remains a key area of focus .