Abstract
The integration of artificial intelligence (AI) in higher education has revolutionized learning practices, yet its impacts on academic competencies remain an area of exploration. This study employed Partial Least Squares Structural Equation Modeling (PLS-SEM), combining reflective and formative measurement models, to investigate the correlation among ChatGPT usage, English academic reading, research skill, and English academic writing. A total of 284 postgraduate students majoring in linguistics and English education across five universities in Indonesia participated. The participants selected through purposive sampling completed an online questionnaire distributed via Google Forms. Findings reveal negative correlation between ChatGPT usage and English academic writing. It implies that ChatGPT alone does not directly enhance the academic writing proficiency of graduate students. The result also shows an unexpected yet significant negative correlation between ChatGPT usage and both English academic reading (β = -0.175, p = 0.017) and research skill (β = -0.164, p = 0.015), suggesting that over-reliance on AI-assisted tools may hinder critical engagement and skill development. Conversely, academic reading (β = 0.640, p = 0.000) demonstrates the strongest influence on academic writing, showing the necessity of cultivating strong reading competencies to enhance scholarly output. These results indicate a pressing need for a balanced approach to AI integration in academic settings, requiring students to actively engage with research materials rather than relying solely on AI-generated content. By fostering academic literacy and research-driven methodologies, educators can mitigate the adverse effects of ChatGPT dependence while maximizing the potential of AI as a supplementary tool. This study contributes to the discourse on AI-driven education, emphasizing the nuanced role of technology in shaping academic literacy and research proficiency.