Generative AI in Higher Education: Conditions for Supporting Students’ Metacognitive Regulation

Purpose of the study. This article aims to identify and systematically analyze the dual effects and risks associated with the use of generative artificial intelligence (GenAI) in higher education, with a specific focus on its impact on students’ metacognitive skills. The study seeks to define the pedagogical conditions under which GenAI functions not as a replacement for cognitive engagement, but as a purposeful instrument for supporting metacognitive regulation – encompassing its three core components: planning, monitoring, and evaluation of one’s own learning. Materials and methods. The study is grounded in a systematic review of empirical research published between 2023 and 2025, including both quantitative and qualitative studies from the fields of pedagogy, cognitive psychology, and educational technologies. Theoretical generalization, synthesis, and structural analysis of scholarly literature were employed in the paper. The theoretical framework draws on J. Flavell’s model of metacognitive regulation and G. Zimmerman’s theory of self-regulated learning, further enriched by insights from distributed cognition theory. Results. The analysis revealed a fundamental duality in GenAI’s influence: on the one hand, it can foster metacognitive development through structured support, socratical interaction, and reflective dialogue; on the other hand, it can provoke such negative phenomena as “metacognitive laziness,” false self-efficacy, and cognitive passivity. Based on the conducted synthesis, an analytical framework is proposed that includes four key conditions for the effective integration of GenAI: (1) explicit definition of the AI’s functional role (coach, opponent, mentor, facilitator, etc.); (2) linking interaction to a specific component of metacognitive regulation; (3) a mandatory reflective component; and (4) teaching prompt engineering as a metacognitive skill. A typology of possible pedagogical strategies is developed, differentiated by three components of regulation: (1) planning – AI as a training coach; (2) monitoring – AI as a “mirror of understanding” or a tool for comparative analysis; and (3) evaluation – AI as a counterargument generator or a facilitator of meta-discussion. Conclusion. Generative AI itself does not carry an unequivocal risk or benefit for metacognitive development; its effect is determined solely by the pedagogical context of integration. The crucial condition for productive use is a reconceptualization of AI’s role – not as a tool for automating thought, but as a dynamic learning partner. The author advocates for cultivating an experimental, reflective, and dialogic culture of GenAI implementation, one centered on preserving and strengthening students’ cognitive autonomy as a foundational goal of contemporary higher education. This approach entails more than mere technical proficiency with AI; it requires fostering students’ critical awareness, readiness for meta-level analysis of their own cognitive strategies, and a sense of responsibility for their learning process.