Reassessing Assessment in EAP

This paper examines how generative artificial intelligence (GenAI) is reshaping assessment design in Australian English for Academic Purposes (EAP) programs. It is motivated by the tension between Australia’s regulatory demands for test-comparable English proficiency evidence and the impact that students’ uses of GenAI has on the validity of EAP assessment tasks. Drawing on an exploratory project with five Master of Applied Linguistics and TESOL interns at Macquarie University College, the study examines students’ experiences in using GenAI tools to complete EAP assignments (literature reviews, annotated bibliographies, research essays, and academic presentations) and considers their validity in high-stakes EAP programs. Findings show that GenAI can lower linguistic and cognitive barriers by supporting source discovery, summarisation, organisation, and language refinement, but also threatens core assessment constructs by obscuring independent language production, evaluative judgement, and task ownership. Across task types, we found vulnerabilities to overgeneralised or misleading summaries, hallucinated or untraceable sources, and growing dependence on AI. Rather than abandoning traditional genres, we recommend redesigning them as process-oriented, AI-aware learning tasks using process artefacts, supervised and oral components, and staged integration of GenAI frameworks.