Inclusive practices mediated by AI and UDL: psychometric evidence and structural models among Italian teachers

Introduction: The incorporation of artificial intelligence (AI) into education opens both opportunities and challenges for equity; when combined with Universal Designfor Learning (UDL), it may strengthen access, representation, and participation. This study examines, in Italian non-university educational settings, how AI-mediated teaching practices aligned with UDL are associated with teachers’ perceptions of inclusion and organisational conditions. Method: A cross-sectional mixed-methods design was employed using a validated questionnaire comprising 29 items (Likert 1–5) and three openended questions. Psychometric validation supported a theoretically coherent threefactor structure consistent with the UDL framework, with adequate reliability (α/ω ≥ .70) and evidence of measurement invariance across educational stage and gender. After informed consent and quality control, 557 teachers participated through a non-probability sampling approach. Results: The results showed substantive associations between familiarity with AI and digital competence (χ 2 = 366.81; V = .47) and between resource availability and frequency of use (χ2 = 76.07; V = .37). Higher levels of agreement were observed for Access/Representation, whereas Self-regulation/Feedback showed the lowest proportion of high agreement. Robust models (HC3) indicated that digital competence and familiarity with AI were consistently associated with higher scores across the subscales. An exploratory structural equation model (SEM) suggested that collaboration and institutional support were indirectly associated with teacher satisfaction through the perceived use of AI aligned with UDL principles. Discussion: These findings suggest that AI and UDL may jointly support inclusive educational practices, depending on individual and organisational conditions.