Google's ATLAS Report: 15M Interactions Reveal Real-World AI Usage

Google and Google DeepMind jointly released the AI & Economy ATLAS v1.0 research report. Based on 15 million anonymized interactions, it comprehensively reveals a map of how real users utilize AI at a large scale. This study fills an empirical gap at the intersection of macroeconomics and AI, serving as a highly valuable, data-driven industry research.

Confirmed

The research data covers multiple product scenarios including Gemini App, Google AI Mode, and Gemini API, spanning over 150 countries and supporting about 140 languages. Using DeepMind's proprietary OCTO tool, these 15 million interactions were accurately mapped to over 800 occupations, 4,000 tasks, and 300 household activities. The report reached a clear quantitative conclusion: 65% of occupations use AI, but it only covers about 21% of specific work tasks. Furthermore, the report confirms that AI diffusion is no longer limited to knowledge work; it is rapidly accelerating into a broader range of real-life scenarios and non-knowledge work.

Why it matters

As noted by @JMateosGarcia, this research provides the most comprehensive picture to date of "how real users use AI at scale," rather than mere product marketing. Stanford researcher @RishiBommasani further pointed out that the report reveals AI adoption in the workplace exhibits a typical pattern of "broad scope (covering many occupations) but shallow depth (used for few tasks)." @danielrock also emphasized that the accelerated spread of AI into non-knowledge work and household scenarios marks a profound penetration into broader economic and life spheres. This holds significant reference value for understanding the actual penetration rate and economic impact of AI across various industries.

2026-07-23 ~ 2026-07-24 · 17 related posts

Primary sources

4 near-duplicate retellings: JMateosGarcia · soumitrashukla9 · HaydnBelfield · minsuk_chang