Harmonised 42-institution public-document coding dataset for visible public generative AI governance in higher education

This dataset supports the manuscript “Visible public generative AI governance in higher education across eight countries”. The dataset contains an institution-level public-document coding dataset for 42 coded public universities within a fixed eight-country sampling frame. The coding captures visible public generative AI governance signals in official university documents, including assessment guidance, academic-integrity linkage, disclosure, privacy and data-handling guidance, ethics or responsible-use guidance, human accountability, curriculum or pedagogy cues, review or update mechanisms, named governance ownership, and staff and student support. The deposit includes the row-level coding dataset, source-log fields, the harmonised codebook, a cross-lineage harmonisation note, and a derived-prevalence check from which the headline figures reported in the manuscript can be recalculated. The dataset is based only on publicly available institutional documents. No human-participant data are included. The original sampling frame specified 80 public universities across eight countries. The deposited dataset reports the 42 institutions for which the coding records and available source-log traceability were harmonised for public deposition. The dataset should therefore be used to recalculate, inspect and reproduce the headline prevalence figures reported for the verified coded corpus, not as a complete row-level source log for all 80 frame institutions. For six rows, artefact notes are preserved but no official URL was retained in the surviving source files; these rows are explicitly marked in the source log using the URL-preservation status field.