Google releases first empirical data on global AI adoption patterns
The tech giant’s inaugural report analyses 15 million interactions to map how artificial intelligence is reshaping tasks across 150 countries.
Google has published the first iteration of its AI & Economy ATLAS (Activity, Task, Landscape, and Adoption Study), marking the launch of a long-term research initiative designed to track the economic impact of artificial intelligence. The report provides an empirical analysis of 15 million aggregated and de-identified human-AI interactions drawn from the Gemini App, AI Mode, and the Gemini API. These platforms collectively serve more than one billion monthly users, offering a dataset that spans over 150 countries, 140 languages, 800 occupations, and 4,000 distinct tasks.
The findings indicate a nuanced shift in workplace dynamics, with most users employing AI to assist with specific tasks rather than fully automating them. Adoption of these tools is widespread across diverse sectors, ranging from office environments to repair shops. This early view of a rapidly evolving landscape suggests that while AI capabilities are advancing, the integration into daily work and life is currently characterised by augmentation rather than replacement.
To process this vast volume of data, the report utilises Google DeepMind’s Observation Clustering and Taxonomy Organisation (OCTO) tool. This technology transforms unstructured text data, such as large language model conversations, into organised entities. The study was developed with stringent privacy protections, including the removal of personally identifiable information and additional layers to eliminate references to sensitive data, ensuring that summaries are aggregated to represent multiple users without linking back to individual logs.
Google acknowledges that ATLAS v1.0 represents only the beginning of a broader research effort and does not capture the full spectrum of economically relevant AI usage. The current dataset excludes interactions with other major products such as Google Workspace, Google Translate, and AI Overviews, as well as enterprise platforms like Gemini for Google Cloud and Gemini Enterprise. It also does not reflect frontier capabilities in areas such as agentic coding and world models, which are expected to be covered as the project expands.
The publication underscores the need for evidence-based research to inform societal decisions regarding AI adoption. Outcomes for the global economy and the nature of work are not automatic, requiring collaboration between researchers, policymakers, businesses, and workers to positively shape the technology’s impact. Contributions to the report were acknowledged from Dame Diane Coyle of Cambridge and Dr David Autor of MIT, as the company continues to refine its methodologies to provide sharper insights into the AI-driven transformation of the economy.
