Answers are grounded in this paper and cite their sources.
Generative AI (GenAI) has been adopted across education faster than its effects on learning have been examined. Critical thinking is among the competencies most consistently demanded of graduates, and whether GenAI use strengthens or weakens it remains unsettled. This review examines how critical thinking is conceptualized and assessed in GenAI-mediated education, what impact GenAI has on students’ critical thinking, and how educators can integrate these tools to support its development. We searched Google Scholar, Semantic Scholar, PubMed, IEEE Xplore, arXiv, and the ACM Digital Library for records on GenAI and education published through June 2026, with no date limit on the conceptual and assessment literature for critical thinking. Supplementary background and pedagogical sources also included institutional reports, assessment resources, books, and educational websites. Records were screened in three stages by title, by abstract and conclusion, and by full text, yielding 66 studies for synthesis. GenAI has no uniform effect on critical thinking. Unstructured use is associated with reduced critical thinking enaction, cognitive offloading, and weakened metacognitive monitoring, while use that requires students to interrogate model output is associated with gains, most consistently in analysis and evaluation. The same models produce both outcomes, indicating that the determining variable is the demand the task places on the student. The configurations that produce gains each make reasoning a condition of task progress. Assessment has not kept pace, since instruments that score the submitted artefact cannot separate student reasoning from model contribution. We propose an 18-item self-report instrument that relocates measurement to the process of engaging with AI across all six dimensions of critical thinking. Whether GenAI strengthens or substitutes for student reasoning depends on task and assessment design rather than on the tools themselves.
No takes yet. Share an insight, caveat, or question.
Shalmani et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: