Generative AI Practices, Literacy, and Divides: An Empirical Analysis in the Italian Context

TL;DR

<3-5 sentence high-level summary> This study provides the first large-scale empirical mapping of GenAI adoption, usage patterns, and literacy in the Italian context. It shows rapid, widespread GenAI uptake and evidence of its substitution of several traditional language technologies, with ChatGPT as the dominant platform. The analysis reveals significant gender and age divides, with literacy in language technologies being a strong predictor of adoption but not fully explaining disparities, indicating other social barriers. The findings highlight the need for targeted education and further research into non-technical factors shaping equitable participation in GenAI technologies.

Abstract

The rise of Artificial Intelligence (AI) language technologies, particularly generative AI (GenAI) chatbots accessible via conversational interfaces, is transforming digital interactions. While these tools hold societal promise, they also risk widening digital divides due to uneven adoption and low awareness of their limitations. This study presents the first comprehensive empirical mapping of GenAI adoption, usage patterns, and literacy in Italy, based on newly collected survey data from 1,906 Italian-speaking adults. Our findings reveal widespread adoption for both work and personal use, including sensitive tasks like emotional support and medical advice. Crucially, GenAI is supplanting other technologies to become a primary information source: this trend persists despite low user digital literacy, posing a risk as users struggle to recognize errors or misinformation. Moreover, we identify a significant gender divide -- particularly pronounced in older generations -- where women are half as likely to adopt GenAI and use it less frequently than men. While we find literacy to be a key predictor of adoption, it only partially explains this disparity, suggesting that other barriers are at play. Overall, our data provide granular insights into the multipurpose usage of GenAI, highlighting the dual need for targeted educational initiatives and further investigation into the underlying barriers to equitable participation that competence alone cannot explain.

Figures (7)

• Figure 1: Language technology adoption and replacement. Panel.a (left) shows the ratio of participants ($n=1,906$) who reported having used each language technology, including GenAI chatbots. Panel.b (right) shows, among users of each technology, the ratio of users reporting that GenAI chatbots have completely or partially replaced their use of that technology. For "Web search", ratios are calculated among GenAI chatbot users ($n = 1,533$). Error bars represent 95% confidence intervals.
• Figure 2: Usage patterns for GenAI chatbots. Panel.a displays frequency distributions across six usage intents with mean frequency scores (0 = never, 1 = less than monthly, 2 = monthly, 3 = weekly). Panel.b shows the total count of activities performed across all users, distinguishing the number of times each activity was indicated for work/study versus personal contexts. Panel.c shows the number of distinct activities each user engages in. The distribution is right-skewed (mean = 5.74 activities). Panel.d displays work/study usage percentage by professional area, with a 50% threshold line (red dashed) distinguishing work/study-dominant from personal-dominant usage. We only include professional area with at least $n=35$. Dots show point estimates, error bars indicate 95% confidence intervals, and sample sizes ($n$) are annotated.
• Figure 3: GenAI Chatbot activity by age groups. Bars show the percentage of users in each age group performing the activities, with 95% confidence intervals. Sample sizes of each age group are in the legend.
• Figure 4: Mean responses on LT literacy, attitudes toward LT education, and bias awareness among GenAI users and non-users. Responses are measured on a five-point scale ranging from $-2$ (strongly disagree) to $+2$ (strongly agree), with $0$ representing a neutral response. We show a restricted range ($-1.5$---$1.5$) for better visualization. Panel.a (left) shows six LT literacy items, with dots representing the group means. Panel.b and Panel.c on the right depict, respectively, the believed importance of establishing LT education in schools and higher education (top) and awareness that LT can reproduce biases (bottom). All differences between chatbot users and non-users are statistically significant according to independent samples t-tests ($p < 0.05$).
• Figure 5: Odds ratios for GenAI chatbot adoption. We show logistic regression results predicting the likelihood of using GenAI chatbots. Points represent odds ratios with 95% confidence intervals. Results are displayed on a logarithmic scale to facilitate comparison of effects across different magnitudes. The dashed line indicates no effect (OR = 1). Reference categories are shown on the right. Statistically significant results are shown as filled markers, while non-significant ones are empty. Results are also available in Table S5 in the Supplementary materials.