Generative AI is being adopted more rapidly than previous digital technologies and has the potential to deliver meaningful productivity gains in professional tasks, raising concerns over whether the benefits are equitably distributed across society. This column analyses Italian household survey data to show that while awareness of generative AI is high, usage is concentrated among younger, more educated and socially engaged individuals, potentially exacerbating inequalities. Targeted interventions such as digital literacy programmes, community-based training, and inclusive design could help bridge gaps, and peer networks and informal learning can also play an important role.

As policymakers deal with the economic and social implications of artificial intelligence, a growing concern is whether the benefits of generative AI (GenAI) are equitably distributed across society. The European Commission’s “Digital Decade” targets aim to ensure that 80% of adults acquire basic digital skills by 2030, yet the rapid diffusion of GenAI tools like ChatGPT raises new questions: Who is using these technologies? For what purposes? And are they translating into tangible economic gains?

Existing evidence indicates that GenAI is being adopted more rapidly than previous digital technologies and has the potential to deliver meaningful productivity gains in professional tasks (Bick et al., 2024). Brynjolfsson et al. (2023) show that GenAI tools can significantly boost productivity in professional writing tasks, while Gambacorta et al. (2024) document a 55% productivity gain in coding tasks for software developers. While having the virtue of being based on experimental analysis, these findings relate to specific group of workers or tasks, providing a narrow picture of the impact of GenAI for the general population and potential societal risks (Managi, 2025). Indeed, less is known about how GenAI is being adopted and used by households, and whether it affects individual earnings or well-being.

In a recent paper (Gambacorta et al. 2025), we contribute to this debate by analysing data from the Italian Survey of Consumer Expectations (ISCE), a nationally representative panel survey that included a dedicated module on GenAI in April 2024. The findings offer a granular view of how Italian households are engaging with GenAI tools, revealing significant disparities in awareness and usage, and modest but meaningful income returns.

Awareness is high, usage is uneven

As of April 2024, 75.6% of Italians aged 18-75 reported being aware of GenAI tools, and 36.7% had used them at least once in the previous year. Monthly usage stood at 20.1%, suggesting that while awareness is widespread, regular engagement remains limited. Prospective usage is more common outside of work – particularly for leisure and educational purposes.

These patterns mirror findings in Aldasoro et al. (2024a), who report similar adoption rates in the US Federal Reserve’s Survey of Consumer Expectations. In both countries, men are significantly more likely to be aware of and use gen AI tools than women, even after controlling for socio-demographic factors. Italian data confirm this gender gap: men are about 5 percentage points more likely to be aware of GenAI and, conditioning on awareness, about 8 percentage points more likely to have used it.

Importantly, this gap persists even when accounting for occupational differences, suggesting that it is not merely a reflection of job roles. Otis et al. (2024) similarly find that men are more likely to use GenAI even when access is equalised, pointing to deeper structural or cultural factors.

Education and age drive adoption

Beyond gender, education and age are the strongest predictors of GenAI awareness and use. Individuals with a college degree are 16 percentage points more likely to be aware of GenAI than those without a high school diploma, and significantly more likely to use it. Younger respondents (aged 18-34) are 11 percentage points more likely to be aware and, conditioning on awareness, 30 percentage points more likely to use GenAI than those aged 65 and older.

These findings echo the ‘digital divide’ documented in other contexts (Doerr et al. 2022), where older individuals perceive fewer benefits from new technologies and are less likely to adopt them. Italian data also indicate that students exhibit the highest levels of awareness and usage, while teachers, though less aware, show high adoption rates once they become familiar with the tools.

Family income also plays a role, albeit a modest one. Higher-income individuals are slightly more likely to be aware of and use GenAI, but the effect is weaker than that of education or age. Living in larger cities has little impact, suggesting that urbanisation is not a major driver of GenAI adoption.

International comparisons: Similar gaps, shared opportunities

The Italian experience with GenAI adoption is not unique. Comparable household-level surveys in other developed economies reveal strikingly similar patterns of awareness and usage, shaped by age, education and gender.

In the US, the Federal Reserve’s Survey of Consumer Expectations included a module on GenAI in early 2024. Aldasoro et al. (2024b) report that awareness of GenAI tools exceeded 80% among adults, with usage rates slightly higher than in Italy. However, the demographic divides were just as pronounced: younger, male, and college-educated individuals were significantly more likely to use GenAI tools.

Across the EU, the ECB’s Consumer Expectations Survey has begun incorporating questions on AI familiarity and use. Preliminary results from the European Commission on European workers suggest that AI awareness is widespread across member states, but usage varies considerably by country, with Northern and Western Europe showing higher engagement than other regions (European Commission 2025).

The World Bank has also explored GenAI usage using unconventional data sources like Google Trends and web traffic (Liu and Wang 2024). Their findings confirm that individual-level engagement with GenAI tools is rising across high-income countries, but access and intensity of use are uneven, often reflecting broader digital divides.

These international comparisons reinforce the Italian findings: GenAI adoption is not just a matter of access, but of capability and relevance. As policymakers consider strategies to promote inclusive digital transformation, understanding these shared patterns can help design interventions that work across borders.

Still modest labour income returns to AI use

Gambacorta et al. (2025) also estimate that GenAI use is associated with a 1.8–2.2% increase in earnings, comparable to half a year of additional education and one-tenth the return to computer use in the early 1990s (Di Nardo and Pischke 1997).

This effect is statistically significant and robust across specifications that include sector and occupation fixed effects. The earnings premium is slightly higher for men, suggesting that GenAI adoption may contribute to a widening gender wage gap. However, we caution that these results are correlational and do not imply causality. Further studies that address the causality issue while expanding the scope of the impact to the entire population of workers are necessary in the future to contribute to the debate.

Conclusion

Generative AI is rapidly becoming a part of everyday life, but its adoption and benefits are unevenly distributed. Our analysis of Italian households shows that while awareness is high, usage is concentrated among younger, more educated and socially engaged individuals. The income returns are modest but meaningful, suggesting that GenAI can enhance productivity without radically transforming earnings.

The uneven adoption GenAI tools raises important policy questions. If younger, more educated, and higher-income individuals are disproportionately benefiting from GenAI, there is a risk that existing inequalities will be exacerbated. Targeted interventions – such as digital literacy programmes, community-based training, and inclusive design – could help bridge these gaps.

Finally, the strong role of social engagement in facilitating adoption points to the importance of peer networks and informal learning. Encouraging community-based experimentation and dialogue around GenAI could be a powerful tool for inclusive innovation.

As the policy debate around AI intensifies, it is crucial to understand how these technologies are being used in practice and by whom. By shedding light on household-level adoption and usage, our research contributes to a more nuanced and inclusive conversation about the future of AI.

Source : VOXeu