Data and code for "Ethical AI guidance in academic libraries and its correspondence with university policies"`

This dataset accompanies a comparative document analysis of ethical AI guidance in academic libraries and its correspondence with university policies. The study coded 40 files from nine universities within a purposive 30-university frame across the United States, United Kingdom, Canada, Australia and China. Twelve ethical dimensions structured separate judgements of coverage, explanation and action support.The current analytical view applies a separately logged six-variable targeted-review overlay to retained historical coding. Among six strict pairs, guidance covered 28 of 44 policy-visible institution–dimension units, supplied explanations for 26 and action support for 24. Historical counts of 30/27/27, original independent codes and reliability estimates are preserved separately. Fifteen topic additions are distinguished by source relationship: nine library-site guides, two reported library-developed teaching themes and four other-unit pages. Topic counts are not counts of independent resources or original educational products.The package supplies current analytical inputs, original coding/reliability copies, source-object mappings, short passage locators, the revision-stage normative-source crosswalk, corpus eligibility records, plotting and verification code, current date-scenario checks, and 52 retained positive-variable witnesses for alternative Queensland assessment versions. A 40-file URL-status manifest contains 27 recoverable stored-page identifiers and 13 explicit NR entries; it is separate from the historical institution-level discovery URLs. These checks introduce no new content coding or reliability estimate.Signed declarations, coder-identifying records, blind-code identity mappings and complete third-party webpages/PDFs are excluded. Short quoted excerpts retain their source rights. File hashes and locators identify captured versions; live content and a precise 30 June snapshot are not reconstructed.