The Effects of Generative AI on Critical Thinking in Educational Contexts: A Systematic Review

Abstract

The rapid integration of generative artificial intelligence (GenAI) tools into educational contexts has intensified debate regarding their impact on higher-order cognitive outcomes, particularly critical thinking (CT). This systematic review synthesizes empirical evidence examining the effects of GenAI on measurable CT outcomes in formal educational settings. Following PRISMA 2020 guidelines, the review searched Web of Science, ScienceDirect, and Google Scholar from database inception to 20th July 2026; 23 peer-reviewed empirical studies met the inclusion criteria. A direction-of-effect synthesis revealed a conditional pattern: most studies reported positive effects when GenAI was embedded in scaffolded, inquiry-driven instructional designs, while a smaller subset reported negative effects associated with cognitive offloading and automation bias; several studies reported mixed or context-dependent outcomes. Effects were moderated by instructional design (scaffolding and prompt engineering), learner prior knowledge and AI literacy, and task complexity. These results suggest that GenAI can support CT when integrated with pedagogical safeguards (verification tasks, formative feedback, and AI-literacy training). Heterogeneity in operational definitions and short intervention durations limits generalisability; we therefore recommend longitudinal RCTs employing validated, performance-based CT measures. This review contributes a mechanism-based interpretation of GenAI’s effects and identifies practical and methodological priorities for research and practice.