Reconsidering AI-detection in EFL writing: a response to De Wilde

Abstract This response engages with De Wilde’s (2024) study on novice EFL teachers’ ability to detect AI-generated writing. While endorsing her call to address AI in teacher education, I argue that her findings also expose wider pedagogical issues. First, novice teachers’ reliance on linguistic deficit cues risks pathologizing proficient learner writing, particularly in contexts where AI is permitted for language editing. Second, treating ‘personal touch’ as a marker of human authorship is increasingly unstable as large language models simulate voice convincingly. Third, drawing on recent teacher-assessment research, I contend that AI detection is a multidimensional, experience-sensitive judgement that extends beyond surface accuracy to rhetorical depth, task engagement, and teachers’ assessment beliefs; exposure to AI texts further shapes detectability. I conclude that teacher training must move from policing authorship toward developing critical AI literacy and ethically integrating AI into process-oriented writing pedagogy.