Speaker
Description
Despite previous studies examining teachers’ AI-detection strategies and measuring detection accuracy and confidence, qualitative evidence on these dimensions remains limited among Vietnamese novice EFL teachers evaluating IELTS-style argumentative essays. This qualitative study investigated the text-based AI-detection strategies, accuracy, and confidence of three novice EFL teachers classifying 15 IELTS Writing Task 2 essays, including 11 human-written and four AI-assisted texts at B2 level or above. Teachers annotated their decisions before semi-structured interviews, and data were analysed using inductive content analysis. The most frequent strategy was examining argument development and organisation, followed by noticing unusually polished language and identifying formulaic and recurrent language patterns; conducting comparative textual checks was least frequent. Although teachers identified nearly all AI-assisted essays, their confidence remained moderate because they considered text-based strategies potentially unreliable in isolation. They, therefore, called for students’ baseline writing samples to support more confident judgments. While the small sample limits the transferability of the findings, the study highlights the need for clearer guidance and training to support novice EFL teachers’ AI-detection judgments.
Keywords: Generative AI; AI-assisted writing; novice EFL teachers; AI-text detection; argumentative writing
