Nearly one in five working-age adults worldwide now uses artificial intelligence tools on a regular basis, according to a global diffusion report published by Microsoft in May 2026. The company’s new metric, called “AI User Share,” pegs adoption at 17.8% of the world’s working-age population as of the first quarter of 2026, up from 12.4% a year earlier. The jump represents roughly 75 million additional users in 12 months, a pace that outstrips early smartphone adoption in several major economies.
The findings land at a moment when governments, employers, and workers are all trying to answer the same question: how fast is AI actually spreading, and who is being left behind?
How Microsoft built the measurement
The methodology behind the metric is detailed in a technical paper titled “Measuring AI Diffusion: A Population-Normalized Metric for Tracking Global AI Usage,” posted on the preprint repository arXiv. Rather than relying on raw web traffic or app downloads, which tend to overcount users in wealthy countries where people own multiple devices, the researchers normalize usage estimates against each country’s working-age population and then apply corrections for device access rates and mobile sharing patterns.
That last adjustment matters more than it might sound. In many lower-income countries, a single smartphone is routinely shared among family members or coworkers. Without accounting for that, a traffic-based count would systematically understate adoption in exactly the regions where growth is fastest.
The denominator for the calculation draws on internet penetration data published by the International Telecommunication Union and hosted by the UN Statistics Division. The ITU collects those figures through annual country-level questionnaires, making them among the most widely cited benchmarks for global connectivity.
What the numbers show, and what they leave out
The 17.8% headline figure, drawn from Microsoft’s accompanying data release, reflects users of web-based and app-based AI products that generate trackable signals: chatbots, coding assistants, image generators, and productivity copilots, among others. It does not capture AI that runs silently inside other software, such as recommendation algorithms in streaming apps or fraud-detection models in banking systems. In other words, the metric measures intentional AI use, not passive exposure.
That distinction matters because it means the real footprint of AI in daily life is almost certainly larger than 17.8%. But it also means the metric is tracking something specific and arguably more meaningful: the share of people who are actively choosing to engage with AI tools.
There are gaps worth flagging. The ITU internet data that feeds the denominator is collected annually, while Microsoft reports AI User Share quarterly. That creates a timing lag: the internet-access baseline for Q1 2026 likely relies on ITU figures from late 2024 or early 2025. In fast-connecting regions like sub-Saharan Africa and South Asia, where tens of millions of people come online each year, the lag could cause the metric to slightly overstate AI’s share of the connected population.
The report also does not break out usage by specific AI platform. Microsoft’s telemetry gives it direct visibility into products like Copilot and Bing Chat, but the company says it incorporates third-party traffic estimates to capture usage of competing tools such as OpenAI’s ChatGPT, Google’s Gemini, and major Chinese-language platforms. Independent researchers have not yet been able to fully replicate those third-party estimates, so the cross-platform numbers carry more uncertainty than the Microsoft-specific ones.
How 17.8% compares to other technology waves
For context, global smartphone penetration crossed the 18% mark around 2013, roughly six years after the iPhone launched. Social media reached a similar share of the world’s adults in about five years. If Microsoft’s numbers hold up under scrutiny, AI tools have reached comparable penetration in roughly three to four years since large language models went mainstream with the release of ChatGPT in late 2022.
The comparison is imperfect. Smartphones and social media required new hardware or new accounts; many AI tools ride on devices and platforms people already own. That lowers the barrier to entry and could explain the faster curve. Still, the speed is striking enough that labor economists and policymakers have taken notice.
The Organisation for Economic Co-operation and Development flagged AI adoption tracking as a priority in its 2025 Digital Economy Outlook, noting that no single metric had yet gained consensus status. Microsoft’s AI User Share is the most detailed public attempt so far, but it is unlikely to be the last. The OECD, the World Bank, and several academic consortia are developing their own frameworks, and the degree to which those converge with Microsoft’s numbers will be a key credibility test in the months ahead.
Where adoption is concentrated, and where it is not
Microsoft’s report shows the sharpest adoption gains in North America, Western Europe, and East Asia, where broadband access, cloud infrastructure, and enterprise software ecosystems create a ready on-ramp for AI tools. The United States, South Korea, and the United Kingdom all show AI User Share figures above 30%, according to the data tables accompanying the methodology paper.
The picture is different in regions where internet access itself remains uneven. Across much of sub-Saharan Africa, AI User Share sits below 5%, a figure that largely mirrors existing gaps in broadband and smartphone access rather than revealing anything new about attitudes toward AI. The metric, in those cases, is measuring digital infrastructure as much as it is measuring AI interest.
That raises a question the report acknowledges but does not fully resolve: can a metric built primarily on web and app telemetry capture the AI tools gaining traction in low-bandwidth environments? Voice-based agricultural advisories in rural India, SMS-driven health bots in East Africa, and offline translation tools used by migrant workers do not always generate the kind of trackable signals Microsoft’s pipeline is designed to detect. If those use cases are undercounted, the global adoption story may be both larger and more geographically distributed than the 17.8% figure suggests.
Why independent replication will determine the metric’s staying power
Microsoft has committed to updating AI User Share on a quarterly basis, with the next release expected in late summer 2026. The company has also published the methodology on arXiv, an open-access repository maintained by Cornell University, inviting outside researchers to stress-test the approach.
That openness is notable for a company with obvious commercial interests in the outcome. Microsoft sells AI products, and a rising adoption curve supports the narrative that its multibillion-dollar investment in OpenAI and its own Copilot suite is paying off. Independent replication of the metric, using non-Microsoft data sources, will be essential before the 17.8% figure can be treated as settled fact rather than a well-constructed estimate.
For businesses planning workforce training or technology budgets, the practical signal is clear even if the precise number is still debatable: AI adoption is no longer a niche phenomenon, and the growth rate over the past year suggests it will not plateau soon. The more urgent question is whether the measurement tools can keep pace with the technology they are trying to track, especially in the parts of the world where AI’s impact on livelihoods may ultimately be greatest.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
