Effective feedback is a central mechanism in second language (L2) writing instruction, shaping learners’ writing development. Recent advances in artificial intelligence (AI) have expanded the scope and immediacy of automated feedback in language learning. This study investigates the role of feedback timing in AI-mediated L2 writing instruction by examining how immediate, delayed, and combined AI-generated feedback influence writing development. Using a quasi-experimental design, this study followed 64 third-year undergraduate learners of Thai (L1 Chinese speakers) over a 12-week classroom intervention. Participants were assigned to one of three conditions: immediate feedback, delayed feedback, and combined feedback. Weekly Thai writing tasks were completed with AI-generated written corrective feedback delivered via ChatGPT, targeting multiple dimensions of writing, including grammatical accuracy, vocabulary use, coherence, and overall text quality. Pre- and post-test writing assessments were used to measure proficiency gains over time. The findings indicate that all three feedback conditions led to significant improvements in L2 writing proficiency. After controlling for baseline differences, both the immediate and combined feedback conditions yielded significantly greater gains than delayed feedback, whereas no statistically significant difference emerged between the immediate and combined conditions. These results suggest that timely, in-session AI feedback supports short-term writing development, whereas combining immediate and delayed feedback does not necessarily yield additive benefits within a single instructional cycle. By empirically isolating feedback timing as a pedagogical variable, this study clarifies when AI-generated feedback is most effective and shows how AI-powered writing tools can be strategically configured to balance real-time linguistic support with reflective revision in classroom-based L2 writing instruction. • AI feedback timing significantly affects L2 writing development • Immediate feedback enhances procedural accuracy and fluency • Delayed feedback supports reflection and deeper processing • Combined feedback integrates procedural and reflective learning • Feedback timing functions as a cognitive learning mechanism
Wiboolyasarin et al. (2026) studied this question.