On Sep 15, 2026, Google published new visualizations and research from its AI & Economy ATLAS, together with a new interactive, open-access experience for exploring millions of global data points. The release pairs occupational adoption data with a study on how scientists are using AI in their work.
Occupational patterns and regional trends
ATLAS data show variation across occupations and countries. In OECD members, computer and mathematical occupations and business and financial operations account for the highest shares of work-related AI use. In non‑OECD countries, office and administrative support; arts, design, entertainment, sports, and media; and educational instruction and library occupations lead adoption.
Google reports that India’s creative industry records 19% of work-related AI usage—about 1.6 times the global average—while the United States leads in technical adoption, with computer and mathematical occupations accounting for 30% of work-related AI usage, roughly double the share in the rest of the world.
Adoption broadly correlates with national income levels, but ATLAS highlights exceptions: Brazil and the UAE show higher AI adoption than their GDP per capita would predict. Use of AI for real-time equipment diagnostics and troubleshooting also varies: Brazil and Germany allocate 7% of work-related AI use to manual tasks (about 1.4× the global average), compared with 4% in Japan.
How scientists are using AI
New research from Google, Google DeepMind, and MIT FutureTech analyzes 2,600 specialized AI models and surveys more than 600 scientists in the U.S. and U.K. The study finds that nearly half of surveyed scientists use some form of AI every day.
The research describes complementary use of LLMs—such as Gemini—and specialized models. LLMs are used across many scientific fields and task types, while specialized models are relatively more common in health and life sciences and for domain-specific prediction, generation, and simulation tasks.
Scientists in the survey report saving just below seven hours per week on average when using AI, which they say frees time for research activities. The study also documents increased time spent validating AI outputs, a growing backlog of hypotheses awaiting tests, and bottlenecks in stages such as physical experimentation and clinical validation. The authors note these factors may limit immediate translation of time savings into new discoveries.
ATLAS is presented as a long-term research project; Google says it will continue collaborating with academic partners to expand analysis and produce further insights into how AI is affecting work and scientific processes.
Original source: Google AI