VOLUNTARY ADOPTION OF GENERATIVE ARTIFICIAL INTELLIGENCE IN HIGHER EDUCATION AND TVET: A DIAGNOSTIC SYSTEMATIC REVIEW OF STUDENT-CENTRED ADOPTION MODELS

Abstract

The rapid diffusion of generative artificial intelligence (GenAI), particularly tools such as ChatGPT, has led to growing voluntary adoption among higher education students. However, within Technical and Vocational Education and Training (TVET) contexts, educational success is ultimately evaluated based on demonstrable competence rather than favourable attitudes or intentions. Despite an expanding body of empirical research on GenAI adoption, existing studies remain predominantly intention-centric, offering limited insight into actual use and sustained engagement that are critical for competence development. This study presents a diagnostic systematic literature review that maps empirical evidence on students’ voluntary adoption of GenAI in higher education and TVET contexts, with explicit attention to outcome boundaries along the intention–use continuum. Guided by PRISMA 2020, a systematic search of Scopus and Web of Science identified 143 records published between 2023 and 2025. Following screening and eligibility assessment, 23 student-only empirical studies were retained for core synthesis. Data were analysed using a descriptive, counting-based evidence-mapping approach focusing on adoption models, construct presence, structural relationships, and outcome categories. The results reveal a strong concentration of studies grounded in UTAUT and UTAUT2 model families, with behavioural intention overwhelmingly positioned as the primary outcome variable. Performance expectancy and social influence consistently emerge as significant predictors of intention, while effort expectancy and facilitating conditions show mixed patterns. Although fewer studies examine downstream outcomes, behavioural intention is generally associated with actual use or continuance when such outcomes are included. However, evidence on sustained and habitual GenAI use remains sparse. By systematically exposing the imbalance between intention-focused and use-oriented evidence, this review contributes a diagnostic empirical map rather than a model extension. The findings underscore the need for future TVET-oriented research to move beyond intention as a proxy for readiness and to prioritise competence-building, repeated, and practice-oriented engagement with GenAI technologies.