Google ATLAS report maps how people use AI across work and daily life

Google has published the first version of its AI & Economy ATLAS, a research effort meant to measure how people are actually using generative AI rather than relying only on surveys, forecasts, or vendor adoption claims. The July 23 post says ATLAS v1.0 is based on 15 million aggregated and de-identified interactions across the Gemini app, AI Mode, and the Gemini API. Google says those products together reach more than 1 billion monthly users, giving the study a unusually broad view of consumer and workplace behavior.

The headline result is that AI use at work appears widespread but still selective. Google reports that workplace AI activity touches every major industry sector and 68% of occupations representing 90% of total U.S. employment. At the same time, in the median job where AI appears, it is used for about 21% of tasks. That supports a more measured picture than predictions in which chatbots rapidly replace whole jobs: people are using AI heavily in some work streams, but not across the full job.

Assistance, not full automation

ATLAS also points to a difference between assistance and automation. Google says most work interactions are collaborative, including ideation, strategy, information retrieval, and learning. Less than 10% of the interactions reviewed fully automate tasks. Non-routine cognitive tasks, such as creative design and hypothesis testing, show up far more often in AI work interactions than they do in the broader economy, suggesting that users are leaning on models where flexible reasoning and iteration matter.

The report is notable for looking beyond office work. Google says workers in manual and technical roles, including auto technicians and industrial mechanics, are using conversational AI for diagnostic and troubleshooting support. In those areas, users are twice as likely to use multimodal AI, such as image or video tools, compared with other work interactions.

Why the dataset matters

Another important finding is that most observed AI use is not workplace use at all. Google says more than 86% of interactions in ATLAS occur outside formal work, including household research, product help, and administrative tasks such as navigating taxes, licensing, and fines. The company argues that those uses may be missed by standard economic measurements, even if they save time for consumers.

Google describes ATLAS as an ongoing project and says the first dataset spans more than 150 countries, 140 languages, 800 occupations, and 4,000 tasks. The company also says the analysis uses privacy protections including removal of personally identifiable information, additional filtering for sensitive content, unlinking from underlying user logs, and aggregation into groups. The result is still a company-authored view into Google's own products, not an independent census of all AI usage. Even so, the scale and task-level framing make it a useful new reference point for policymakers, researchers, and businesses trying to understand where AI adoption is concrete and where it remains uneven.

Sources

• blog.google
• ai.google

Cover photo by Keysi Estrada on Pexels, used under the Pexels License.