AI-assisted Teaching-learning Materials for School Students: A Critical Narrative Review of Opportunities, Classroom Practices and Implementation Challenges

Abstract

Artificial intelligence (AI) is reshaping how teaching-learning materials are designed, adapted, delivered and evaluated in school education. Earlier AI-supported materials relied mainly on constrained intelligent tutoring, adaptive practice and automated feedback, whereas contemporary generative AI can produce explanations, examples, questions, simulations, code, images and dialogic support in response to natural-language prompts. This critical narrative review examines how these capabilities can contribute to school students' learning, the practices through which teachers and students are using AI-assisted materials, and the conditions under which benefits or harms are more likely. Literature was selected through iterative searches of education-focused and multidisciplinary scholarly indexes, DOI and publisher records, and authoritative institutional sources, with emphasis on peer-reviewed K-12 research and evidence available through 27 June 2026. The synthesis indicates that AI-assisted materials can improve the responsiveness of instruction by supporting differentiated explanations, formative feedback, self-regulated learning, inquiry, programming and teacher preparation. Yet positive effects are not uniform. Studies vary in design quality, duration, subject area and degree of teacher mediation; some report gains in achievement or engagement, whereas others find weaker flow, self-efficacy or learning than conventional support. Generative systems also introduce epistemic risks through inaccurate or untraceable outputs, cognitive risks through over-reliance, and governance risks concerning privacy, bias, copyright, assessment integrity and unequal access. The evidence therefore supports a pedagogy-first model in which AI augments rather than substitutes for teacher judgement, materials are grounded in curricular sources, outputs are verified, students are taught to interrogate AI responses, and use is age-appropriate and governed by clear data and accountability rules. Future research requires longer-term, independently replicated studies across primary and secondary settings, stronger measures of learning processes and equity, and direct comparisons among design features rather than between AI access and non-access alone.