Designing AI-Aware Assessment in Higher Education: A Cross-National Analysis of Generative AI Guidance and Design Patterns in English-Medium Universities

Generative AI is reshaping coursework and assessment in higher education, yet institutional guidance still varies in how clearly broad principles are translated into assessment practice. Guided by pattern-based educational design, the study used a framework-based document analysis of 21 official, public-facing institutional and teaching-unit guidance documents from universities in Europe, North America, Australasia, and Asia. The analysis combined directed qualitative coding with descriptive cross-case summaries of permitted and prohibited uses, disclosure and verification requirements, ethical safeguards, and ratings of clarity, actionability, and assessment specificity. Assessment-related configurations were synthesised into recurrent design patterns, and a descriptive ethical gap map compared the presence or absence of provisions related to privacy, consent, retention, fairness, and accessibility. Across the corpus, all universities allowed some student use of generative AI under conditions, and almost all required disclosure. Policies were generally clear, actionable, and assessment-specific. Four patterns were especially prominent: (1) process-evidence portfolios with AI-use logs, (2) AI-assisted drafting plus in-class or oral verification, (3) critical engagement with AI outputs, and (4) secure high-stakes exams combined with AI-enabled coursework. The ethical mapping showed stronger attention to public-upload risks and bias awareness than to data retention, consent language, and accessibility alternatives. Overall, the findings suggest that universities in this purposive sample have moved beyond blanket prohibition toward more assessment-aware guidance, while important governance and equity gaps remain. The resulting pattern library is offered as a provisional, policy-grounded resource for translating high-level AI principles into concrete assessment design options.