Effective Methods of Integrating ChatGPT into Student Learning for Achievement: A Systematic Review and Meta-Analysis

• In Event: Generative AI in Education: From Hype to Evidence-Based Practice

Abstract

ChatGPT, powered by generative AI, is increasingly used to support student learning. However, its impact on academic achievement and optimal integration methods remain underexplored. This meta-analysis synthesizes 44 empirical studies using a random-effects model to estimate the overall effect, classify integration approaches, and compare their effectiveness. Results show a small-to-moderate improvement in academic outcomes (Hedges’ g = 0.418). A 2 × 2 classification framework identified five integration categories, explaining substantial heterogeneity. Meta-regression evaluated different integration modes and revealed three high-impact conditions: ChatGPT spans all learning phases, active teacher guidance, and theory-based interventions. These findings provide a timely, practice‑oriented summary of how ChatGPT is currently being integrated into student learning and what works best under varying conditions.

Authors

• Jia Chen Liu, The Chinese University of Hong Kong

• Alan C. K. Cheung, The Chinese University of Hong Kong