The intersection of internet epistemic beliefs and perception of writing in shaping students’ views of ChatGPT’s feedback

This study addresses the limited research on how students’ internet epistemic beliefs (IEB) and perceptions of writing (PoW) jointly influence their engagement with AI-generated feedback, particularly across English as a Foreign Language (EFL) and English as a Second Language (ESL) contexts. A sequential explanatory mixed-methods design was adopted, involving 235 undergraduate students from Indonesia (EFL) and Malaysia (ESL). Data were collected through a structured questionnaire measuring IEB, PoW and perception of ChatGPT (PoC), followed by semi-structured interviews that explored students’ epistemic stance and experiences with ChatGPT in writing tasks. Structural equation modeling (SEM) showed that IEB strongly predicted PoC (β = 0.827, t = 15.5, p < 0.001), whereas PoW had no significant effect (β = −0.0658, t = 1.23, p = 0.219). The model explained 61.8% of the variance in PoC. Thematic analysis further revealed students’ epistemic vigilance, pragmatic reliance on ChatGPT for linguistic and organizational support, and their consistent preference for human feedback in evaluating content quality. This finding suggests that while students’ beliefs about online knowledge critically shape their evaluation of artificial intelligence (AI)-generated feedback, their writing-related motivation and self-efficacy may not directly determine whether they perceive such feedback as useful. Pedagogically, this highlights the need to integrate epistemic literacy into writing instruction through practical strategies such as source triangulation, peer-AI feedback comparison, and reflective tasks on the reliability of AI outputs.