Abstract
<p>Generative AI (GenAI) has diffused rapidly across Europe, yet who feels able touse it—and who is willing to work with it—remains poorly understood. Using thelatest CRONOS-3 panel of the probability-based European Social Survey (N =7,556; 11 European countries, fielded in 2025), we model six separate measuresof GenAI literacy and use frequency, together with acceptance of GenAI as anassistant and as a peer in the workplace, using country-clustered estimators withwild-bootstrap inference. Acceptance is strongly role-differentiated: a majorityaccept GenAI as an assistant, but only one in ten as a peer. Both competence andacceptance are sharply stratified by education, age, and gender—and the genderdivide is the most consistent pattern of all: women trail men on every one of thesix literacy and use measures and on both acceptance outcomes, and the deficitappears in behavioural use as well as in self-assessed skill, so it cannot be reducedto gendered self-assessment alone. GenAI literacy and use frequency in turnstrongly predict acceptance in both roles. Cross-national variation is substantial,but with only 11 countries no association between national media, scholarly, orAI-talent environments and individual outcomes survives strict small-cluster andmultiple-testing corrections, so national context is treated descriptively. GenAIcompetence and workplace acceptance thus emerge as socially stratified from theoutset—with women consistently furthest behind—an early-forming divide thatrisks hardening as GenAI becomes workplace infrastructure.</p>