Many fifth-grade students in Indonesia still write poems with limited diction and imagery, and little is known about how generative artificial intelligence can be embedded in contextual pedagogy for young learners. This study examined whether integrating ChatGPT into Contextual Teaching and Learning (CTL) improves fifth-grade students' poetry writing, creative thinking, and language literacy compared with conventional CTL. A quasi-experimental non-equivalent control group design was applied to 55 fifth-grade students in Bima City, Indonesia (experimental n = 28; control n = 27) over four learning sessions. Rubric-based tests covering seven indicators (diction, imagery, rhyme and rhythm, ideational fluency, originality, vocabulary, and literary appreciation) were calibrated with the Rasch model (person reliability = 0.93). Data were analysed using independent-samples t-tests, normalised gain (N-gain), and effect sizes. The experimental group achieved a moderate mean N-gain (55.10%) compared with a low gain in the control group (14.51%; Welch's t = 4.24, p < .001, d = 1.15). However, posttest differences between the groups were not statistically significant for any of the seven indicators (p = .088–.363), and no meaningful advantage was found for creative thinking or language literacy. ChatGPT-assisted CTL therefore shows promise for supporting gains in poetry writing, but the evidence is preliminary given the small sample and short intervention. Longer, adequately powered studies with ethical safeguards for young AI users are recommended.
Copyrights © 2025