GenAI-agents in education: a systematic review of macro-level trends, educational roles, theoretical foundations and future agendas

Abstract

With the rapid advancement of Generative AI (GenAI), autonomous agents are increasingly integrated into education, yet research remains fragmented. Guided by the Educational Transformation Scenarios Framework, this review examines GenAI-agents in education across three interconnected levels: macro, meso, and micro. This study conducts a systematic literature review following PRISMA guidelines, analyzing 33 empirical studies (2022–2025) from Wos, IEEE, and Scopus. Findings reveal a surge in research since 2022, primarily in higher education using quasi-experimental designs. The study identifies four core educational agent roles: Cognitive-Epistemic, Self-Regulatory Support, Affective Support, and Human-AI Interaction Agents. These roles respectively foster higher-order thinking, self-regulated learning, socio-emotional competencies, and AI literacy. Furthermore, the theoretical foundations are synthesized into three domains: Cognitive and Learning, Motivational and Affective, and Pedagogical Knowledge and Technology Adoption. While GenAI-agents demonstrate multi-role collaborative potential, results suggest that future research must address trust, governance, contextual adaptation, and multi-agent ecosystems. By integrating existing empirical evidence, this review provides a comprehensive framework for understanding the educational value of GenAI-agents and offers a roadmap for their sustainable development within educational systems.