GenAI Disclosure in University EFL Writing: A Policy Discourse Analysis (2023–2025)

Abstract

Generative artificial intelligence (GenAI) has entered the tertiary writing ecology with unprecedented speed, fundamentally challenging long-standing assumptions about academic integrity, authorship, and the legitimacy of writing assistance. This study investigates how universities construct “acceptable use” of GenAI within policy discourse, with a specific focus on disclosure—what students are required to declare, when disclosure is triggered, and how it is proceduralized. Employing a qualitative Policy Discourse Analysis (PDA) framework, the research examines a cross-national corpus of official, university wide instruments published or updated between 2023 and December 2025. The corpus includes academic integrity frameworks, assessment regulations, and institutional GenAI guidance from six research intensive institutions: King Abdulaziz University, King Saud University, and Qassim University (Saudi Arabia), alongside the University of Toronto (Canada), and the University of Sydney and UNSW Sydney (Australia). Findings indicate a significant divergence in institutionalization. International cases increasingly adopt structured “lane based” or “level based” frameworks that routinize disclosure as a metacognitive practice linked to assessment design, whereas Saudi institutions exhibit a rapid, tech forward adoption discourse aligned with Vision 2030 but a more variable picture of proceduralized disclosure at the university wide level, with some cases characterized by documented policy silence. The analysis foregrounds procedural justice risks for multilingual writers in English as a Foreign Language (EFL) contexts, who face a “surveillance tax” due to systemic biases in automated detection tools. The study argues that disclosure functions as a vital visibility technology that renders AI mediated assistance legible, reduces reliance on forensic suspicion, and enhances fairness for language diverse writers, and it concludes with actionable recommendations for standardizing disclosure triggers and role descriptions to support valid assessment in the GenAI era.