Governing generative AI in digital education: how institutional guidance becomes course-level policy

Abstract

As generative artificial intelligence becomes embedded in digital education, universities face a governance problem that institutional guidance alone cannot resolve: how should acceptable AI use be defined within particular courses and assessments? This study examines syllabi as formal governance texts through which university principles become student-facing rules. Using qualitative comparative document analysis, the study analyzed 35 syllabi collected from a large U.S. public research university, including 22 from Education and 13 from other fields. Course-level statements were compared with the institutional governance framework. The analysis distinguished primary governance stances and examined institutional alignment, policy rationales, discourse registers, and variation across course contexts. The institutional framework delegated substantial authority to instructors while emphasizing communication, attribution, verification, and student responsibility. Course-level enactment was highly heterogeneous: 11 syllabi were silent on student AI use, 11 were prohibitive, five permitted specified uses, two broadly permitted AI with responsibility safeguards, and six treated AI or machine learning as an object of pedagogical or professional learning. Six syllabi were explicitly aligned with institutional guidance, seven implicitly aligned, eight elaborated the institutional framework, three provided minimal guidance, and 11 remained silent; none directly contradicted a specific institutional requirement. Authentic or independently produced work was the most common rationale, appearing in 18 of 24 non-silent syllabi. Governance patterns crossed disciplinary boundaries, while all pedagogical cases were concentrated in AI-adjacent courses. The findings challenge simple disciplinary explanations and show that course-level AI governance is shaped by the interaction of assessment design, authorship expectations, course purpose, AI adjacency, and instructor discretion. Variation is best understood as an outcome of delegated digital governance whose educational value depends on clarity, justification, and alignment with the intellectual work students are expected to perform.