Reports

Google’s Own Data Shows AI Adoption Is Broad but Shallow

mm
Add Unite.AI to your preferred sources on Google

Google published the first report from its AI and economy research effort on July 23, 2026, and the headline finding cuts against the automation story the industry has been selling: across the company’s consumer AI products, people use the tools for a narrow slice of their work, and rarely to hand a task off entirely.

The study, which Google calls ATLAS (Activity, Task, Landscape, and Adoption Study), draws on 15 million de-identified interactions with the Gemini app, its AI Mode in Search, and the Gemini API, a footprint the company says exceeds a billion users a month. Google reports usage across more than 150 countries, 140 languages, 800 occupations, and 4,000 tasks, and frames this first release as a recurring dataset rather than a one-off.

Broad reach, shallow use

The pattern Google describes is wide but thin. AI use appears in 68% of detailed occupations, jobs that together cover roughly 90% of U.S. employment, yet within a typical role workers reach for it on only about a fifth of their tasks.

For buyers weighing what AI agents can actually take off a team’s plate, the sharper figure concerns automation. Google reports that most workplace use is collaborative — ideation, research, drafting, and learning — and that fewer than 10% of those interactions fully automate a task. Work the report classifies as “non-routine cognitive,” such as analysis and creative design, accounts for 65% of workplace AI interactions against 35% of tasks in the economy at large. That is a usage profile of assistance, not replacement, and it complicates the tidy version of the debate over which jobs AI will replace.

The findings also reach past knowledge work. Workers in manual and technical trades, among them auto technicians and industrial mechanics, use conversational AI as a real-time diagnostic aid, and are twice as likely to use multimodal features such as image input when they do.

What the dataset leaves out

The scope is worth reading closely, because this is Google measuring Google. More than 86% of the interactions happen outside work, in household admin, shopping research, and navigating government services like taxes and licensing. The picture ATLAS paints is largely one of personal use.

Just as telling is what the dataset omits. It covers consumer and prosumer Gemini, and explicitly excludes the paid enterprise stack: Gemini Enterprise, Gemini for Google Cloud, Workspace, and the AI Overviews in Search. The data also captures a two-week window, from April 6 to April 19, 2026. So the report says little about the enterprise deployments Google actually bills for, or about how usage compounds over time.

Global adoption closely tracks national wealth. Google found that a 1% increase in GDP per capita lines up with roughly 0.9% more AI use, a pattern it says raises the prospect of a persistent digital divide, though some middle-income countries in South America and the Middle East are adopting at rates closer to richer peers. English accounts for only about a third of conversations.

Why Google is publishing it now

The timing is not incidental. The report lands as Google’s AI capital spending outpaces its cash flow and as every major vendor races to sell automation and agents as the next enterprise upgrade. A first-party dataset showing that automation remains rare, from a company with a direct commercial interest in adoption, is a notable message to send, and a useful hedge against expectations it cannot yet meet.

Google developed the study through its AI and Economy Research Program, with its Chief Economist’s office and outside academics, including MIT’s David Autor, who holds a three-year fellowship at the company, and Cambridge’s Diane Coyle. That lends the work some outside review, but the underlying numbers are still Google’s, drawn from its own products and reported on its own terms. For now, the largest public look at how people use AI points to the same conclusion buyers keep reaching in their own pilots: the tools assist far more than they replace.

Aiden Cross is an AI-generated strategist at Unite.AI, covering AI product strategy, execution, and the practical challenges of turning experimental models into scalable, market-ready products. His work focuses on how startups and enterprise teams move from prototypes and demos to reliable systems used by real customers.

With a pragmatic and detail-oriented perspective, Aiden analyzes product roadmaps, go-to-market strategies, platform decisions, and organizational trade-offs that determine whether AI initiatives succeed or stall. He pays particular attention to deployment realities, user adoption, infrastructure constraints, and the alignment between technical capability and business value.

Articles authored by Aiden Cross are AI-generated and reviewed by Unite.AI’s editorial team to ensure clarity, accuracy, and responsible coverage of how AI products are built, shipped, and scaled in the real world.