Abstract

This study investigates Chinese university students’ acceptance of AI-generated feedback in academic English writing, focusing on DeepSeek, a domestically developed multilingual platform. Drawing on an extended Technology Acceptance Model (TAM) that integrates Perceived Usefulness (PU), Perceived Ease of Use (PEOU), Feedback Clarity (FC), and Trust (T), the research examined how these factors predict students’ Behavioral Intention (BI) to continue using DeepSeek for AI-generated feedback. Data were collected through a structured questionnaire administered to 293 non-English major undergraduates across two universities. Descriptive, correlational, and regression analyses revealed that students generally held positive perceptions of DeepSeek (M = 4.05–4.26), with Feedback Clarity emerging as the strongest predictor of continued use (β = 0.583, p < .001), followed by Perceived Usefulness (β = 0.118, p = .023). Correlation analysis further indicated significant positive relationships among all variables (r = 0.480–0.736, all p < .01). Although Trust and Perceived Ease of Use were positively correlated with Behavioral Intention, they did not exert significant effects in the multivariate model. These findings suggest that clarity and pedagogical relevance outweigh technical ease once baseline functionality is assured. The results support the applicability of the extended TAM framework in the context of EFL writing and highlight the importance of designing AI feedback systems that provide clear, actionable, and trustworthy guidance. Pedagogically, the study underscores the need to integrate AI tools purposefully into writing instruction to enhance feedback accessibility and student engagement. These insights contribute to the growing scholarship on AI-mediated learning in higher education.