From generative to agentic AI: qualitative thematic synthesis of cognitive and motivational impacts on personalized higher education pathways

This qualitative thematic synthesis examines how generative and agentic AI reshape cognitive and motivational processes in personalized higher-education pathways. Drawing on a purposively sampled corpus of 33 systematically coded studies via a three-stage thematic synthesis framework (Thomas & Harden, Citation2008), the review generates three analytical themes: an Agency Shift in which AI transitions from reactive tool to proactive learning partner; a Cognitive Trade-Off where performance gains coexist with metacognitive risk; and an Autonomy Paradox in which personalization simultaneously empowers and constrains learner self-direction. Evidence was appraised through a three-layer quality assurance framework encompassing per-paper evidence richness assessment, synthesis-level rigor assurance (composite score 0.890), and finding-level confidence evaluation using GRADE-CERQual (Lewin et al., Citation2018). The synthesis proposes a four-level typology of AI roles – reactive generative assistant, adaptive feedback system, semi-agentic learning planner, and fully agentic pathway orchestrator – and identifies discipline-specific and learner-heterogeneity boundary conditions. Findings suggest that educational depth depends less on AI sophistication than on coordinated design of prompt literacy, assessment practice, and integrity governance within interactive learning environments.