Microsoft's AI report highlights uneven adoption across countries

According to Microsoft's AI Diffusion Report 2025, adoption is growing much faster in wealthier countries, widening the gap between the Global North and Global South.

By the end of 2025, generative AI had moved from a niche tool to something many people now use in daily life. Around 16.3 percent of the global population had used a generative AI product, up from 15.1 percent just months earlier. That move may appear small on paper, but it implies that one in six people worldwide now uses AI for learning, work, or problem-solving.

The data comes from Microsoft’s AI Diffusion Report 2025, which looks at how many people actually used generative AI tools during the period studied. Rather than tracking interest or downloads, the report focuses on real usage. It relies on aggregated and anonymized telemetry data and accounts for differences in internet availability, device ownership, and population size to prevent overstating adoption in more connected regions.

Even with those adjustments, the divide between regions remains hard to ignore.

A widening gap between regions

Use of generative AI grew much faster in wealthier economies than in developing ones. By the end of the year, close to 25 percent of the working-age population in the Global North had used a generative AI tool. In the Global South, the figure stood just above 14 percent.

That gap did not narrow over time. It grew again in the second half of the year, suggesting that access, infrastructure, and policy choices continue to shape who benefits first. Growth is happening almost everywhere, but not at the same speed.

Countries with the highest adoption rates seem to follow a familiar pattern. They invested early in digital infrastructure, taught basic AI skills to their workers, and encouraged its usage in public services. These steps did not guarantee rapid growth, but they lowered friction when generative AI tools became widely available.

The United Arab Emirates and Singapore remain well ahead of the global average. In 2025, about 64 percent of the UAE’s working-age population used generative AI, compared with just under 61 percent in Singapore. Both countries continued to grow, though at a slower pace than before.

Several European countries followed behind. Norway, Ireland, France, and Spain all recorded gains, but their growth rates suggest adoption may be starting to settle among early leaders rather than accelerate further.

Technical strength does not guarantee public use

The United States stands apart from many of these trends. It remains central to AI research, infrastructure, and model development, yet public use lags behind several smaller economies. By the end of 2025, around 28 percent of working-age Americans had used generative AI, placing the country 24th worldwide.

The numbers point to a broader pattern. Strong private-sector development and research capacity do not automatically translate into everyday use. In several cases, countries with fewer technical resources moved faster in turning AI tools into something people regularly use at work or at home.

Most countries held relatively stable positions in the rankings. South Korea was a clear exception.

In just a few months, it climbed from 25th to 18th place. Generative AI use rose from about 26 percent of the population to more than 30 percent, well above the global growth rate. South Korea also became the second-largest market for ChatGPT subscriptions, behind only the United States.

That jump did not come from a single factor. It reflected a mix of policy decisions, technical progress, and public interest.

Policy clarity and language support

Government action played an important role. In late 2025, South Korea formalized its AI governance through the National AI Strategy Committee and passed the AI Basic Act. These steps provided clearer rules for regulation and public-sector use. At the same time, AI education programs grew beyond major cities, strengthening ties between universities and regional institutions.

Progress in language support was also important. For years, several large language models performed poorly in Korean. That changed in 2025, when newer models demonstrated significant improvements. Benchmark results increased from basic comprehension to levels closer to those of top university students. For Korean speakers, AI tools became practical for writing, translation, education, and professional tasks in ways they had not been before.

A cultural occasion boosted attention. In April 2025, Ghibli-style graphics generated with ChatGPT spread swiftly on Korean social media. Many first-time users were drawn to the feature’s ease of use. Usage data showed that some of them continued to use AI technologies even after the trend passed.

Open models reach overlooked regions

South Korea’s rise reflects what can happen when language support and public policy align. Elsewhere, a different pattern emerged.

DeepSeek, an open-source AI platform, has found popularity in places where many Western services are limited or expensive. By releasing its model under an MIT license and providing a free chatbot, it lowered both cost and technical barriers.

Adoption stayed modest in North America and Europe, but grew quickly in China, Russia, Iran, Cuba, Belarus, and across parts of Africa. In several African markets, usage levels are estimated to be two to four times higher than in many Western countries.

What adoption depends on next

DeepSeek’s spread highlights a broader point. AI adoption depends on more than model performance. Access, language support, pricing, and political context all shape who uses these tools and how quickly. Open models can move fast where commercial platforms face limits, though that speed also raises questions about oversight and standards.

By the end of 2025, the overall picture is clear. Generative AI use is rising, but not evenly. Countries that invest early and set clear rules are pulling further ahead. Others remain behind, even when they have strong technical foundations.