<![CDATA[Purpose of the study: This study systematically reviews and synthesises the empirical literature on the role of Artificial Intelligence (AI) in science education, focusing specifically on how distinct AI features support the phases of students’ self-regulated learning (SRL) and the construction of conceptual understanding in science. Methodology: A qualitative systematic literature review reported in line with PRISMA principles was combined with bibliometric mapping. Peer-reviewed articles were retrieved from the Scopus and Sinta databases (2020–2026). Metadata and keyword co-occurrence were analysed using VOSviewer, followed by a thematic synthesis of 25 eligible articles structured around an AI-feature × SRL-phase × conceptual-understanding extraction matrix. Main Findings: AI research in science education has grown sharply since 2022. Intelligent tutoring systems, learning analytics, adaptive simulations, and generative AI predominantly scaffold the forethought and performance phases of SRL, enhance conceptual understanding, reduce misconceptions, and raise motivation through personalised interaction. Support for the self-reflection phase remains comparatively weak, evidence is concentrated in higher education, and longitudinal and developing-country studies are scarce. Novelty/Originality: The review advances an integrative framework that explicitly links specific AI affordances to individual SRL phases and to conceptual change in science, repositioning AI as a cognitive and metacognitive mediator rather than a content-delivery aid, and derives an evidence-based agenda that foregrounds the reflection phase, K–12 science, and Global South contexts.]]>
Copyrights © 2026
