Google launches global study of millions of AI chats to understand how people use artificial intelligence
Through its "ATLAS" effort, the major tech company Google is working to analyze how people are using its artificial intelligence products.
Google has opened a window into how its AI products are actually being used. The company released the first dataset from its AI & Economy ATLAS initiative, drawing on 15 million de-identified interactions across the Gemini App, AI Mode, and the Gemini API. The scale is significant: more than one billion people touch these tools each month, and the dataset spans 150 countries, 140 languages, 800 occupations, and 4,000 tasks.
The findings quietly puncture several narratives. Over 86 percent of conversations happen outside formal work. Inside the workplace, AI adoption covers roughly 88 percent of U.S. employment, but penetration is shallow and overwhelmingly collaborative. End-to-end task automation remains limited. The implication is straightforward: AI is everywhere in the office, but it is assisting rather than replacing.
The language data tells a parallel story. English accounts for only about a third of global conversations, and users do not appear to be abandoning their native languages for complex professional tasks. Work and non-work usage show nearly identical language distributions, suggesting that AI is being integrated into existing linguistic habits rather than reshaping them.
The strategic logic is worth noting. Google controls one of the largest AI user bases on the planet, and it is now publishing the most granular dataset on real-world AI behavior to date. This positions the company as the default reference point for researchers, policymakers, and competitors trying to understand adoption patterns. The data is also a competitive instrument: by defining what "comprehensive" looks like, Google sets the benchmark others must meet.
For the labor market, the signal is measured. AI is woven into a wide range of occupations, but the depth of use remains thin. The technology is augmenting tasks rather than displacing workers at scale, at least for now. The ATLAS dataset will give employers, educators, and analysts a clearer baseline for tracking whether that balance shifts over time.