Enhancing EFL Students’ Reading Comprehension through ChatGPT-Assisted Instruction: A Quasi-Experimental Study

Abstract

The use of ChatGPT in English language learning has created opportunities for providing interactive and individualized instructional support. However, empirical evidence regarding its effectiveness in improving the reading comprehension of junior high school EFL students in Indonesia remains limited. This study aimed to examine the effectiveness of ChatGPT-assisted instruction in improving the reading comprehension of eighth-grade students at SMPN 1 Paiton. A quantitative approach with a quasi-experimental design was employed. The participants consisted of 54 eighth-grade students, with 28 students in the experimental group and 26 students in the control group. Data were collected through pre-tests and post-tests consisting of 20 multiple-choice reading comprehension items. The data were analyzed using descriptive statistics, the Shapiro–Wilk normality test, Levene’s homogeneity test, paired sample t-tests, an independent sample t-test, and N-Gain analysis. The results showed that the experimental group improved from a pre-test mean of 50.96 to a post-test mean of 85.77, while the control group improved from 49.46 to 72.14. The paired sample t-test showed significant improvement within both groups (p < 0.05). The experimental group also achieved a higher N-Gain score of 72.71, categorized as high, compared with 46.37, categorized as medium, in the control group. However, the independent sample t-test yielded p = 0.117, indicating that the post-test difference between the groups was not statistically significant at the 0.05 level. These findings suggest that ChatGPT-assisted instruction was associated with greater descriptive improvement in reading comprehension, although further research with larger samples and longer interventions is needed.