Background/purpose: The present quantitative study investigated the effectiveness of utilizing ChatGPT as an instructional tool to enhance the learners' grammatical proficiency. Materials/methods: The study involved 60 Omani intermediate EFL learners randomly divided into three groups with 20 learners. The control group received traditional face-to-face instruction. Experimental group 1 received in-class instruction supplemented by ChatGPT, which provided feedback on grammatical structures. Experimental Group 2 engaged in fully online instruction, with ChatGPT as the primary feedback facilitator for grammatical tasks. Some researcher-made grammar tests were designed and piloted before the study to ensure the reliability and validity of the assessment tools. Results: Statistical analysis indicated that all groups improved grammatical proficiency from the pretest to the posttest. However, the progress observed in Experimental Group 1 was statistically more significant. The comparison of the pretest and delayed posttest scores revealed a sustained improvement in the scores of Experimental Group 1, indicating the long-term efficacy of combining in-class instruction with ChatGPT feedback. In contrast, the scores of both the Control Group and Experimental Group 2 showed a notable decline from the posttest to the delayed posttest, suggesting limited retention of grammatical knowledge in these groups. Conclusion: ChatGPT makes it possible to give particular feedback that adapts to the process, enhancing motivation and facilitating learning in various ways. The research validates the educational advantages of ChatGPT but recommends additional studies on its usefulness for working with various populations of students, various types of grammar as well as the quality of students and their teachers' attitudes and abilities to utilize technology.