Who wrote this? An EAP think-aloud study on AI detection

Abstract

Generative AI (GenAI) is a rising trend in Higher Education (HE); it is not only reshaping how students learn but also how academic integrity is upheld. This is particularly evident in English for Academic Purposes (EAP), where authentic academic writing is a core indicator of learning outcomes. Despite the growing use of AI detection tools, their documented unreliability leaves educators facing uncertainty and ethical risk when grading their students’ writings, as any misclassification could lead to serious consequences. This exploratory study examines how eight foundation-level EAP teachers in Bahraini HE institutions approach the task of detecting AI-generated content. Using a think-aloud protocol and Cognitive Engagement as an analytical lens, the study reveals that teachers rely on a blended approach, shifting flexibly between minds-on strategies (contextual knowledge, linguistic analysis, professional reflection, and verification via interrogation) and hands-on techniques (use of external tools), which indicates that AI detection is a complex professional judgment task rather than a purely technical one. The results suggest that Professional Development (PD) should go beyond general AI awareness and focus on practical, discipline-specific skills, such as interpreting unreliable detector results, identifying changes in students’ writing patterns, and dealing with uncertainty when evidence is not sufficient.