Exploring EFL Learners’ Integration and Perceptions of ChatGPT's Text Revisions: A Three-Stage Writing Task Study

ChatGPT can promptly reformulate a text and improve its quality in content and form while preserving the original meaning. Yet, little is known about how learners respond to such reformulations. Here, we employed a three-stage writing task (composing-comparison-rewriting) to investigate how learners notice, integrate, and perceive ChatGPT's reformulations as feedback in English as a foreign language (EFL) classroom context. We collected learners’ written notes made during the comparison of their original texts and the reformulations, categorizing them based on the language-related type of noticing (vocabulary, discourse, and form) and quality of noticing (depth of processing: low, intermediate, and high). We also analyzed the instances of reformulations integrated into learners’ rewriting and their answers to a questionnaire. The results showed that: 1) the reformulations directed learners’ attention to the gaps in their original writing, especially in word choice, and prompted them to integrate ChatGPT-generated changes in their rewriting; 2) the number of instances integrated into the rewriting was directly related to the quantity, type, and quality of noticing in the comparison stage; and 3) learners generally appreciated the pedagogical value of ChatGPT in EFL writing, particularly during the revision stage, although they believe occasionally ChatGPT might misinterpret their intentions. The study suggests that ChatGPT's reformulations should be complemented with peer and teacher feedback to create a comprehensive and personalized learning environment.