Why Chunks Go Unnoticed: A Longitudinal Study of a ChatGPT-Aided Chunking Approach in EFL Classrooms

Lexical chunks are essential for fluent language use, yet many EFL learners struggle to notice them in authentic input. This study investigated the effectiveness of a ChatGPT-aided chunking approach designed to enhance learners’ awareness of lexical chunks through explicit instruction and contextualized noticing activities. Seventy first-year Chinese EFL learners participated in a 13-week intervention integrated into a College English course, with complete data obtained from 62 students for statistical analysis. Using a one-group pretest–posttest design, learners completed chunk-recognition tasks before and after the intervention, followed by a language proficiency test and a perception questionnaire. Results showed a significant increase in the mean number of chunks recognized, from 11.39 (SD = 7.19) in the pretest to 20.81 (SD = 8.17) in the posttest, t(61) = −11.96, p < .001. Chunk recognition ability was also positively correlated with overall English proficiency, r(60) = .618, p < .001. Questionnaire results indicated that learners perceived the approach positively and reported improvements in multiple language skills. Despite these gains, chunk recognition remained relatively limited, suggesting that noticing lexical chunks is a more challenging process than is often assumed. The findings highlight the importance of explicitly fostering chunk awareness and suggest that generative AI can serve as a valuable tool for supporting chunk-based instruction in EFL classrooms.