The rapid diffusion of generative artificial intelligence (AI) into schooling raises urgent questions about whether “AI literacy” functions as an equalizing resource or is absorbed into existing stratifications. Using Bourdieu’s concepts of field, capital, and habitus as an interpretive lens, this study investigates how AI-related practices are defined, acquired, policed, and converted into advantage across twelve K–12 schools (public, private, and international) in one Egyptian governorate. The study adopts a comparative embedded case design drawing on 411 semi-structured interviews with students, teachers, and administrators, 75 institutional documents, and non-participant observations. A mechanism-oriented thematic analysis shows that AI literacy does not operate as a single, uniform competence. Instead, three recurrent and contested AI literacies circulate across settings—AI-as-shortcut/cheating literacy, AI-as-ethics discourse literacy, and AI-as-technical production literacy—whose legitimacy and exchange value depend on institutional mechanisms. Exam-security and surveillance regimes most strongly shape what becomes publicly legible as “risk,” frequently producing concealment dispositions and disqualifying shortcut practices. Recognition infrastructures (e.g., rubrics, portfolios, awards, and counselling narratives) determine whether technical competence becomes institutionalized capital or remains “talent without exchange value.” Linguistic/curricular alignment and shadow learning ecologies further generate within-type stratification, including self-taught production trajectories that remain non-convertible where recognition pathways are absent. Overall, AI literacy in this context functions as fractured cultural capital: it is variably treated as risky/illegitimate, reputationally valuable, or professionally convertible, in ways that tend to reproduce rather than offset educational inequality. The study is bounded to one governorate and includes a single elite international case; findings are therefore field-sensitive tendencies rather than nationally generalizable claims.
Mekheimer et al. (Fri,) studied this question.