Pandu Prasodjo, Siusana Kweldju, Nur Mukminatien, Francisca Maria Ivone, Hilarius Raditya Priambada Purba
This study examined the influence of Google Gemini, a generative artificial-intelligence (AI) tool, on the lexical appropriateness of Indonesian EFL students’ paragraph writing. Drawing on theories of lexical richness and appropriateness, the study employed an exploratory quasi-experimental design involving three writing conditions: a baseline task, an AI-assisted revision task, and a delayed post-test without AI support. Fourteen second-year English Language Teaching students (A2–B1 CEFR) participated in the study. Lexical appropriateness was measured using Type-Token Ratio (TTR) and Measure of Textual Lexical Diversity (MTLD), and the data were analyzed using repeated-measures ANOVA to examine differences across tasks. The results revealed no statistically significant differences in lexical diversity across the three conditions, suggesting that short-term interaction with Google Gemini did not lead to measurable improvement in lexical appropriateness. Qualitative observations further indicated that limited vocabulary depth and low metalinguistic awareness constrained students ’ ability to evaluate and apply AI-generated lexical suggestions. These findings suggest that lexical appropriateness is reflective, and instruction-dependent process, and that AI tools such as Google Gemini function more as a potential scaffolding partner rather than autonomous enhancers of lexical development. © The Authors 2026.
Universitas Negeri Malang, Jl. Cakrawala No.5, Sumbersari, Kec. Lowokwaru, Jawa Timur, Kota Malang, 65145, Indonesia; Universitas Internasional Batam, Baloi-Sei Ladi, Jl. Gajah Mada, Tiban Indah, Kec. Sekupang, Kepulauan Riau, Kota Batam, 29426, Indonesia
