Generative AI Adoption in Higher Education: Mapping Institutional Policy Responses and Student Usage Trends from Public University Data and National Surveys

Abstract

Generative artificial intelligence (GenAI) has moved from novelty to routine study tool in under four years, yet university governance has followed a slower and more uneven path. This study combines two sources of publicly available evidence to map that gap. First, it codes the official GenAI guidance of a purposive sample of 40 highly ranked universities across six world regions (39 analysable) on eight governance dimensions, with a blind second coding of a random subsample (pooled Cohen's kappa = 0.71). Second, it synthesises repeated national and international surveys of students and staff from the United Kingdom, the United States, Australia and multinational samples published between 2024 and 2026. Central guidance was found at 34 of 39 universities (87%), but only 8 used a formal policy instrument. Half of those with guidance (17 of 34) delegated GenAI rules to individual instructors without an institutional default, 7 permitted use by default with acknowledgement and 6 prohibited it by default. Disclosure of use was required at 23 (68%), and 19 (56%) discouraged or disabled AI text detection. A composite governance index (0 to 8) was higher in North America, Europe and Oceania (mean 6.64) than in Asia, Africa, the Middle East and Latin America (mean 4.07; Mann-Whitney p = .006). Among institutions that had published guidance, however, the regional difference was small and not significant (6.64 vs 6.11; p = .25), so the gap lies mainly in whether guidance is public rather than in its design. Survey data show UK undergraduate use rising from 66% in 2024 to 95% in 2026, and use in assessed work from 53% to 94%, while only 5% (2024) and 25% (2026) of students in global samples were fully aware of comprehensive institutional guidelines. The findings point to a policy-practice gap: institutions regulate mainly through instructor discretion and disclosure, while students have already normalised GenAI in assessed work.