Applications of artificial intelligence (AI) in academic writing have introduced heightened levels of technological monitoring, particularly affecting students' emotional states within educational environments. This study employed a quantitative research design to examine factors predicting academic anxiety among a sample of 300 Chinese pre-service teachers who frequently utilise AI-assisted tools such as ChatGPT and Grammarly. The investigation explored how concerns regarding false detection, policy ambiguity, perceived academic pressure, and awareness of AI detection contribute to academic anxiety, and how these factors subsequently influence academic confidence, using Structural Equation Modelling (SEM) for analysis. The findings reveal that the most significant contributors to elevated anxiety are the fear of being incorrectly identified and the lack of clarity in institutional policies, both of which negatively affect academic confidence. Perceived academic pressure also demonstrated a positive, moderate association with anxiety, whereas awareness of AI detection showed a weaker yet notable effect. These outcomes underscore the potential psychological consequences of AI surveillance in educational settings, emphasise the importance of clarifying institutional policies, and highlight the role of digital literacy in mitigating academic stress and enhancing students’ self-efficacy. The study carries important policy implications concerning the balance between technological advancement, student mental wellbeing, and academic integrity, thereby supporting the development of teacher education programmes.