AI usage reaches 18.8% worldwide as the regional gap widens

Nearly one in five working-age people worldwide used a generative AI product in June 2026, according to a new Microsoft report. Its estimate is 18.8%, up about one percentage point from the first quarter. The increase spread across almost every economy in the company's dataset, but the distance between higher- and lower-adoption regions widened rather than narrowed.

Microsoft published the second-quarter update on September 21. The headline number is an estimate built from anonymized product telemetry and adjustments, not a global census of AI users. That distinction matters when comparing countries or inferring what people actually do with the tools.

The gap behind a rising average

The report puts AI usage at 28.8% in the Global North and 16.2% in the Global South. Both figures rose, but Microsoft says their gap continued to widen. It also reports unusually high adoption in the United Arab Emirates, at 73.3%, and Singapore, at 64.3%.

South Korea recorded the largest absolute quarter-on-quarter increase, 3.5 percentage points. Saudi Arabia moved five places to 25th in Microsoft's rankings. Japan rose by 2.2 percentage points from its first-quarter level of 22.5%, about a 10% relative increase. These numbers describe the report's particular measurement, rather than an agreed count of every AI interaction in those countries.

The contrast is more revealing than the leaderboard. A global average can climb while access diverges. Differences in connectivity, devices, skills, language support and cost can determine whether a new product is usable in ordinary work or study, even where interest in it is strong.

What Microsoft's meter can and cannot see

The company defines diffusion as the share of people aged 15 to 64 who used a generative AI product during the measured period. It derives estimates from aggregated, anonymized Microsoft telemetry, then adjusts for operating-system and device market share, internet access and population. The method attempts to make the numbers comparable across economies, but it remains a model whose coverage depends on the products and signals it observes.

Microsoft acknowledges that newer AI tools and models have grown beyond its current measure. It says the next report will expand coverage, likely lifting estimated user share in almost every economy and especially in China. That planned change may complicate comparisons with earlier quarters. Readers should look for a clear bridge between the old and revised series when those figures arrive.

The report also uses consumer Copilot insights to examine how AI is used. It finds a higher share of education and learning activity in the Global South. That observation comes from Microsoft's own product, so it should not automatically be generalized to all AI services or classrooms.

A possible route around access costs

Open-weight models are taking a growing share of token volume in API usage, Microsoft says. They may make local-language adaptation and cheaper access easier, particularly where commercial subscriptions or foreign-language interfaces are obstacles. Infrastructure, connectivity and training still determine whether that promise turns into broad use.

The next meaningful test is not simply whether the 18.8% rises. It is whether the gap between regions narrows under a broader measurement and whether adoption supports useful work, research and learning. Microsoft's figures show momentum and unequal reach at the same time.