Skip to content
Review Open access

Generative Artificial Intelligence in High School Science Education: A Systematic Review of Learning Outcomes, Pedagogical Strategies, and Critical Thinking Development

Nov 2026 · Journal of Education and Learning Reviews · 0 citations · 52 references

Abstract

Background and Aim: Evidence on generative artificial intelligence (AI) in secondary science education remains limited. This systematic review examined students’ conceptual understanding and engagement, teachers’ pedagogical strategies and challenges, and students’ critical thinking and scientific inquiry. Materials and Methods: Following PRISMA 2020, ERIC, Scopus, and Lens.org were searched for peer-reviewed empirical studies published between 2022 and 2026. Two authors independently screened 187 unique records (Cohen’s κ = .78), identifying 20 eligible studies. Methodological quality was assessed using the Mixed Methods Appraisal Tool. Framework synthesis integrated findings, informed by Technological Pedagogical Content Knowledge (TPACK). Results: Seven studies assessed AI capabilities rather than classroom implementation. Only two compared learning outcomes across conditions; none employed a randomized controlled design. Student perceptions and engagement were generally positive, whereas learning outcomes were mixed. A chatbot supporting self-regulated learning outperformed a rule-based chatbot, but ChatGPT-4o yielded smaller gains than peer collaboration. Teachers reported concerns regarding content accuracy, assessment reliability, and instructional readiness. Generative AI matched or exceeded student performance on structured examinations but demonstrated lower grading reliability than human assessors. Evidence concerning critical thinking was particularly limited. Richer reasoning was reported with dialogue-oriented tools or teacher guidance, whereas answer-seeking use of general-purpose tools was associated with weaker reasoning. Conclusion: Pedagogical structuring was recurrently associated with productive AI use, although causal inference remains limited. A provisional model warrants empirical testing. Priorities include structured tasks, science-specific teacher development, human assessment oversight, explicit AI literacy instruction, school governance, and controlled comparisons of structured and unstructured use across secondary science classrooms.

Read PDF

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.