Examining the Impact of Preparation Difficulties on Postgraduate Entrance Exam Progress Among Foreign Language Students in the AI Era: A PLS-SEM Study

Generative AI has changed the job market that foreign language (FL) graduates in China face, and many now treat postgraduate study as a way to improve their prospects. What they struggle with while preparing for the entrance examination, and how those difficulties combine to slow preparation down, is less well understood, especially at local universities. Drawing on career decision-making difficulty theory and research on test anxiety, this study models five dimensions of preparation difficulty against perceived preparation progress. Sixty final-year FL students preparing for the 2027 National Postgraduate Entrance Examination completed the survey. Partial least squares structural equation modeling in SmartPLS 4 was used, with dimension merging and iterative indicator pruning as robustness checks. Only academic and psychological difficulties (β = −0.258, p = .008) and time and self-discipline difficulties (β = −0.349, p = .015) predicted progress. The refined model kept 20 indicators across five dimensions (composite reliability ≥ 0.73, all HTMT ratios below 0.85), and both significant paths held in the robustness specification. Open-ended responses pointed to senior-year course loads and administrative duties, which suggests that reducing coursework and protecting study time matter more than information campaigns or financial aid on their own.