<![CDATA[Artificial intelligence (AI) is increasingly integrated into science education as instructional support and a partner in dialogue. This Scopus-based systematic review and thematic meta-synthesis examined AI roles, pedagogical design, human–AI interaction, reported learning outcomes, and implementation conditions in literature dated 2023–2026. From 237 identified database records, 49 science-education reports met the substantive inclusion criteria; two additional plant-science reports were documented as contextual boundary cases but excluded from the core synthesis. The synthesis identified AI roles as a tutor, scaffold, content generator, dialogic partner, adaptive system, multimodal resource, and analytical tool. Selected empirical interventions reported gains in achievement, problem posing, retention, transfer, or self-efficacy, whereas evidence for critical thinking and problem solving was mixed or limited. Teacher capacity, pedagogical design, interaction quality, institutional readiness, and ethical safeguards shaped the interpretation of these outcomes. Science educators should therefore use AI to support—not replace—learners’ scientific reasoning, verification, and teachers’ professional judgment.]]>
Copyrights © 2026
