Mind the Gap: AI Adoption in Europe and the U.S.

For three decades, American workers have been pulling ahead of their European counterparts. Between 1995 and 2025, output per hour increased by 88% in the U.S. versus just 30% in the 20 countries using the euro.In 2026, Bulgaria became the 21st country to use the euro, but it is not included in this data, which is through 2025. Prior research has linked this widening gap to greater diffusion of information and communication technologies (ICT) in the U.S. American firms invested more in computers and software, adopted them faster and more effectively, and reaped larger productivity gains as a result.

Now, a new wave of technology is arriving. Advances in artificial intelligence (AI) have the potential to reshape work across many sectors of the economy. As with the ICT revolution, the economic impact of AI will depend critically on how quickly and broadly workers and firms adopt it. Will the U.S. again pull ahead, or will Europe be able to close the gap?

To find out, we combined evidence from worker and firm surveys conducted across the U.S. and Europe in 2025 and 2026. Our findings, published in a new paper prepared for the Brookings Papers on Economic Activity Spring 2026 Conference,The working draft (PDF) of this paper was presented at the Brookings Papers on Economic Activity (BPEA) Spring 2026 Conference. The final version will be published in the spring 2026 volume of the BPEA journal. Conference drafts, slides and presentation recordings are available on the Brookings website. point to the following conclusions:

• AI adoption is substantially higher in the U.S. than it is in Europe, though there is wide variation across European countries. More broadly, richer countries adopt AI at higher rates.
• Industries with higher AI adoption have experienced faster productivity growth, both in Europe and the U.S. As of now, we do not find evidence that AI adoption is associated with job losses at the industry level.

How Did We Measure AI Adoption?

Our analysis drew on two main sources of data.

For workers, we extended our U.S. Real-Time Population Survey (RPS) to cover six European countries: the four largest European economies (Germany, the U.K., France and Italy) and two leading digital economies (Sweden and the Netherlands). The RPS has tracked generative AI adoption among U.S. workers since June 2024.Results from these earlier surveys are discussed in a September 2024 On the Economy blog post and this November 2025 follow-up. However, we surveyed both U.S. and European workers in May-June 2025 and again in January-February 2026. The surveys were designed to be nationally representative of employed workers in each country.

For firms, we employed the EU’s survey on ICT Usage and E-Commerce in Enterprises (EU-ICT-Firm), conducted annually by Eurostat and national statistical agencies. The 2025 wave covered roughly 157,000 firms across 32 European countries. Firms were asked whether they use any of eight specific AI technologies, such as machine learning, speech recognition, natural language generation and image recognition. This broad coverage gives us an unusually comprehensive picture of firm-level AI adoption across the continent.For the U.S., we have information on firm-level AI adoption from the Business Trends and Outlook Survey (BTOS), administered by the U.S. Census Bureau. We draw on the BTOS in the underlying paper but do not use it in this blog post because differences in how the BTOS and EU-ICT-Firm survey ask about AI adoption make direct comparisons challenging. A future blog post will discuss these measurement differences in detail.

What Did We Find?

Finding 1: The U.S. Leads Europe in AI Adoption, but There Is Wide Variation within Europe

The figure below shows that 43.0% of U.S. workers reported using AI for their jobs in January-February 2026, which is the highest share among the seven countries we surveyed. AI adoption in European countries ranged from 36.3% in the U.K. down to 25.6% in Italy, meaning that U.S. adoption was between 18% and 68% higher than in any of these individual European countries.

Comparing the share of workers who use AI is only part of the picture. Our surveys also asked workers how many hours per day they spend using AI and how many days they use it during the week. When we combined these answers, we found that 5.2% of all U.S. work hours in early 2026 were spent using AI, roughly double the share of those in the U.K., Sweden and the Netherlands, and more than triple the share of those in Germany, France and Italy. The gap in work hours spent using AI is wider than the gap in worker adoption because, beyond being more likely to use AI in the first place, U.S. workers use it more intensively.

Finding 2: Richer Countries Adopt AI at Higher Rates, and Worker and Firm Adoption Move Together

The worker and firm data tell a consistent story: Across country-industry pairs, the correlation between worker and firm AI adoption is 0.75, suggesting that the patterns we are seeing are real and economywide, not a quirk of how any single survey asks about technology use.

Like worker adoption, firm adoption varies widely across countries. Among the 32 European countries covered by the EU-ICT-Firm survey, the share of firms using at least one AI technology averaged 20%, ranging from more than 35% in Denmark, Finland and Sweden to less than 10% in Cyprus, Greece, Bulgaria, Poland, Turkey and Romania. The next figure provides a simple way of summarizing this pattern: Richer countries adopt AI at higher rates. This is not a new story. It mirrors the spread of earlier technologies like computers and the internet: Wealthier nations adopted them first, with broader diffusion following over time. The open question is whether those lagging will catch up, and if so, how long it will take. We show in our paper that, since 2023, early adoption leaders have tended to pull further ahead of those adopting AI at lower rates—at least for now.

Finding 3: Industries with Higher AI Adoption Have Experienced Faster Productivity Growth, but Not a Drop in Employment

The ultimate question, of course, is whether any of this matters for the economy. Does AI adoption actually translate into faster productivity growth? Do we already see effects on employment?

The following figure offers suggestive evidence that, in the U.S., it does impact productivity. The figure plots industry-level AI adoption rates against annualized productivity growth relative to each industry’s pre-COVID-19-pandemic trend, from the fourth quarter of 2022 through the third quarter of 2025. The pattern is clear: Industries in which more workers use AI have grown faster, on average, than their pre-2020 trajectories would have predicted. Quantitatively, a 10 percentage point increase in AI adoption among workers is associated with 2.9 percentage points of additional cumulative productivity growth over this period. Parallel analyses using cross-country and cross-industry variation in firm AI adoption across European nations point in the same direction, as do results for a longer window starting in 2019 for both the U.S. and Europe.

We want to be clear about what these estimates can and cannot tell us. They are not causal. AI adoption may be highest in industries that were already on a productivity upswing relative to trend for other reasons, and the time period during which AI has been more widely adopted by workers is short. But the fact that similar patterns emerge when using worker adoption data for the U.S. and firm-level adoption data for Europe as well as different empirical approaches makes a purely coincidental explanation harder to sustain.

How large could the aggregate effect be? As of 2026, 43% of U.S. workers used AI for their jobs, compared with an average of 32% across the European countries we surveyed, a gap of 11 percentage points. Scaling our estimates to that gap suggests 3.2 percentage points of additional cumulative productivity growth in the U.S. relative to Europe since 2022.

Another big question looming over the economic impact of AI is whether the technology will replace workers. We conducted the same exercise for employment as we did for productivity. We found no clear evidence that AI adoption was associated with either job gains or job losses at the industry level, in Europe or the U.S., as can be seen in the final figure. While this may change as the technology matures, the early data offer little support for the most alarming scenarios about AI-driven job displacement.

Conclusion

History offers a cautionary tale. Since the 1990s, greater ICT investment by the U.S. helped propel rapid productivity gains, while European productivity fell behind. Our findings show that the U.S. is building a similar lead today in AI adoption, and we provide suggestive, though not conclusive, evidence that AI may already be affecting aggregate productivity.

Notes

1. In 2026, Bulgaria became the 21st country to use the euro, but it is not included in this data, which is through 2025.
2. The working draft (PDF) of this paper was presented at the Brookings Papers on Economic Activity (BPEA) Spring 2026 Conference. The final version will be published in the spring 2026 volume of the BPEA journal. Conference drafts, slides and presentation recordings are available on the Brookings website.
3. Results from these earlier surveys are discussed in a September 2024 On the Economy blog post and this November 2025 follow-up.
4. For the U.S., we have information on firm-level AI adoption from the Business Trends and Outlook Survey (BTOS), administered by the U.S. Census Bureau. We draw on the BTOS in the underlying paper but do not use it in this blog post because differences in how the BTOS and EU-ICT-Firm survey ask about AI adoption make direct comparisons challenging. A future blog post will discuss these measurement differences in detail.