Meta-analysis: The Effect of AI Tools on EFL Learners’ Writing Performance

다국어 초록 (Multilingual Abstract)

This meta-analysis examines the effects of generative AI and AIassisted writing tools on EFL
learners’ writing performance. Nine empirical studies published between 2023 and 2025 were
analyzed, comprising 15 independent groups and a total of 744 EFL learners. Effect sizes were
calculated from pre–post writing scores using a random-effects model to account for
methodological variability. The results indicate that AI-integrated instruction leads to
significantly greater overall writing gains than traditional instruction. Subgroup analyses reveal
strong positive effects on language use, content, organization, and vocabulary, while fluency
shows a small negative trend. These findings suggest that AI tools primarily enhance writing
quality through improved accuracy and reflective revision rather than increased writing speed.
Overall, the study highlights the value of AI tools as a complementary instructional tool when
integrated with appropriate pedagogical guidance in EFL writing instruction.

Keywords: EFL learner, EFL writing, AI-assisted writing, generative AI,
writing proficiency, meta-analysis

목차 (Table of Contents)

• CHAPTER 1: INTRODUCTION 1
• CHAPTER 2: LITERATURE REVIEW 3
• 2.1. Overview of AI tools in Language Education 3
• 2.2. Learning Theories 5
• 2.3. AI in EFL Writing: Potentials and Limitations 6
• CHAPTER 3: METHOD 8
• 3.1. Research Design 8
• 3.2. Data Collection and Study Selection 9
• 3.3. Coding Framework and Data Extraction 14
• 3.4. Effect Size Calculation and Statistical Analysis 15
• CHAPTER 4: RESULTS 17
• 4.1. Overall Effectiveness of AI Tools (RQ1) 18
• 4.2. Differential Effects by Writing Domain (RQ2) 20
• 4.3. Verification of Results 23
• 4.4. Summary of Findings 25
• CHAPTER 5: Discussion 26
• 5.1 Interpretation of Findings 26
• 5.2 Limitations 27
• 5.3 Future Research and Implications 28
• CHAPTER 6: Conclusion 29
• REFERENCES 31
• APPENDIX 36
• Appendix A. Final List of Included Studies and Excluded
• Studies 36
• Appendix B. Coding Data for Overall Analysis 37
• Appendix C. Coding Data for Subgroup Analysis. 38
• Appendix D: R Code for Meta-analysis 38