<![CDATA[This study investigates the influence of AI assistance on Theme choices and Thematic Progression (TP) patterns in Indonesian EFL students’ discussion texts. A qualitative comparative discourse analysis was employed using a paired-text design in which the same 20 participants produced one text independently (pre-AI) and another with ChatGPT assistance (post-AI) by direct within-subject comparison of thematic change. A total of 558 clauses were analyzed (265 pre-AI; 293 post-AI) at the clause and discourse levels using Systemic Functional Linguistics frameworks. Results indicate that AI assistance produced a notable increase in the use of unmarked topical themes +6.7%, and a corresponding decrease in the Marked Topical Themes -7.1%, while Textual Themes showed only a slight increase +0.4% and the Interpersonal Themes remained mostly stable -0.7%. Regarding TP patterns, Derived/Split Progression showed the most significant change +2.8%, while Constant/Parallel Progression decreased -3.4% and Linear/Zigzag remained stable -0.2%. The findings suggests that AI supports structural organization and coherence in EFL writing through standardization of clause structure and modest thematic elaboration, while having limited impact on interpersonal or evaluative expression. This study provides pedagogical implications for combining AI tool with explicit writing instruction in order to enhance discourse organization, thematic development, and coherence.]]>
Copyrights © 2026
