This study examined how learners’ use of ChatGPT (GPT-4) during a revision task was associated with changes in second language (L2) writing performance and explored its relationship with learners’ level of language analytic ability (LAA), as well as their perceptions of the revision process. Twenty-six Korean university students completed IELTS Academic Writing Task 2 essays before and after a ChatGPT-supported revision activity. Writing performance was evaluated in terms of task response, coherence and cohesion, lexical resource, and grammatical range and accuracy. LAA was measured using an artificial-language rule-induction test, and learner perceptions were collected through surveys and interviews. Results showed gains across all assessed dimensions following the revision activity. Correlational analyses indicated that LAA was associated with improvements in grammatical accuracy and task response, but not with lexical resource. Learners generally perceived ChatGPT as helpful for identifying issues and supporting revision, while also noting the cognitive effort required to engage with its feedback. These findings suggest that ChatGPT can function as a useful tool for supporting L2 writing revision; however, the extent to which learners benefit from such support may depend, in part, on their language analytic ability.

I. Introduction
II. Theoretical Background
    1. AI-generated WCF: Learner Outcomes and Perceptions
    2. WCF and Language Analytic Ability
III. Research Method
    1. Context and Participants
    2. Instruments
    3. Procedure
    4. Data Analyses
IV. Results
V. Discussion
    1. Writing Development Following ChatGPT-Mediated Revision
    2. LAA as a Moderator of ChatGPT-Generated WCF Uptake
    3. Learner Perceptions of ChatGPT-Mediated Revision
VI. Conclusion
References
APPENDICES
    Appendix 1: Pretest and Posttest Essay Topics
    Appendix 2: LAA Test (Sample)
    Appendix 3: Questionnaires for Learner Perceptions
    Appendix 4: ChatGPT Guideline

• Ji Hyun Kim(Keimyung University)