The Maze: Generative AI appears to have a rich-world adoption problem. Monthly use reaches 45.0% of the US population but only 29.2% worldwide, a 15.8-point gap. Change the denominator, and the story flips. Among people already online, 48.8% in the US and 46.1% worldwide use generative AI at least monthly. The difference is just 2.7 points. AI appetite is already global. Internet access is not.
The headline gap is mostly a connectivity gap. The source population split separates monthly generative-AI users, online non-users and offline people. In the US, 45.0% use generative AI, 47.2% are online but do not, and 7.8% remain offline. Worldwide, the shares are 29.2%, 34.1% and 36.8%. The offline-population difference is 29.0 points—almost twice the 15.8-point difference in overall AI adoption. That is the denominator doing the heavy lifting.
Connected people behave far more alike than national averages imply. Monthly adoption among internet users is 48.8% in the US and 46.1% worldwide. Put differently, nearly half of connected people on both sides of the comparison have entered at least one prompt during the month. Lower population-wide penetration does not automatically mean weak interest, cultural resistance or a failed product. It can simply mean the serviceable market stops where the network does.
The addressable market still has billions of people outside it. The ITU estimates that 6 billion people, or 74% of humanity, were online in 2025, leaving 2.2 billion offline. That is not a niche distribution problem. It is a market boundary. Every new connection can add a potential AI user, shopper, seller or digital-service customer before any platform wins another point of preference among people already connected.
Access opens the door; capability decides the value. The World Bank warns that connectivity alone does not guarantee broad economic gains. Affordability, devices, bandwidth, digital skills, language support and institutions shape what happens next. The source counts anyone who enters one LLM prompt per month, so it cannot distinguish a casual experiment on a free tier from daily professional use with advanced tools. Participation may converge while outcomes remain uneven.
Why it matters: AI companies and commerce operators should separate demand from distribution. In lower-connectivity markets, weak population penetration may say more about internet reach than product appetite. That changes the playbook. Lightweight interfaces, low-end device support, local languages, affordable data and trusted onboarding can expand the market faster than another premium feature. But connectivity is only the first gate. Winning the next billion users means making AI reachable, useful and economically relevant after they arrive. That distinction matters for every digital business forecasting demand beyond today’s connected customer base.
Sources: EMARKETER source post | ITU Statistics | World Bank Atlas
