Preliminary research dataset: first 40 institutions Version: 0.2-preliminary Date: 2026-10-19 Status: Preliminary This is a preliminary research dataset, not a study. It is not representative and supports no population conclusions, category frequency comparisons, rankings, or grades. The minimum of 400 coded institutions has not been met. Inter-rater reliability and Cohen's kappa have not been established. There has been no second human blind coding and no independent human second review. 40 institutions; 11 variables; 440 cells; 79 coded values; 361 unknown; 12 institutions with zero coded values. These are file counts, not findings. Internal source SHA-256: ec88f05f7a901db413ab50b08b2ada724f6b14d22098798eb788f942f32c4f57 FILES university-ai-policy-first-40-preliminary.csv Columns: order_rank, ipeds_unitid, name, state, control, level, size_band, enrollment_fall2023, ai_default_stance, ai_policy_level, ai_in_misconduct_definition, ai_disclosure_expected, ai_detector_stance, plagiarism_defined, sanction_structure, appeal_process, honor_code_type, writing_center_linked, detection_software_named_by_institution, fields_coded, policy_urls_reviewed, accessed_on, dataset_status, notes university-ai-policy-first-40-preliminary-fields.csv Columns: order_rank, ipeds_unitid, name, variable, value, source_url, verified_date, evidence_quote, source_kind, http_status university-ai-policy-codebook.csv Columns: variable, allowed_categories, definition, export_note university-ai-policy-first-40-README.txt: this documentation. The first 40 institutions follow a fixed, randomly ordered sampling sequence. The frame uses the IPEDS HD2024 directory and EF2023A enrollments downloaded on 2026-10-03. Order SHA-256: 529aad6d2da16cea20e6c392e5b5ee6a9f61de24344870c60fb8102111fcec90. The six searches cover honor code, academic integrity, plagiarism, generative AI, ChatGPT, and artificial intelligence policy. Only public pages are used. Sources include official policy pages, handbooks, catalogs, and institutional guidance. Every coded field has a source URL, a verification date, and a short evidence quote. The codebook calls for quotes under 25 words. Unknown means no defensible value was found within our search protocol, not that an institution has no policy. The protocol checks official institutional pages, uses six site: searches, and has a 15-minute time box. Codebook v1 categories named not_found are exported as unknown and do not mean no policy. Unknown rows in the long file have empty source columns. Evidence and retained notes are copied without rewriting. Notes containing internal workflow terminology are omitted. Coverage is limited to public pages. Login-only policies may be missed. Campus or school applicability is partly unresolved; school-scoped statements must not be generalized. Policies can change, and AI extraction errors are possible. Institutional identity and enrollment reflect the frozen frame, not a new check of current institutional status. CONTRIBUTORS Research and initial coding were performed by Claude-based AI research agents in Elitewerk OS using public websites. A second AI agent checked a sample of the first version: this was AI QA, not human review. A targeted AI follow-up on 19 October 2026 rechecked sources for ten risk cases and made 19 field changes at four institutions (ranks 6, 23, 29, 35): 7 previously unknown fields were coded and 12 source or procedural details were updated. Other proposed additions, including additions for two further institutions, were deliberately not included. Markus Nick, responsible for this page's content and research, states that he personally looked through the data and performed his own quality control as of 19 October 2026. There is no row-by-row review log, no second human coder, and no external review. CORRECTIONS kontakt@manimedia.de Institutional statements are reported neutrally, without judging institutions. CHANGELOG v0.2 preliminary, 19 Oct 2026: first public release. A targeted AI follow-up on 19 October 2026 rechecked sources for ten risk cases and made 19 field changes at four institutions (ranks 6, 23, 29, 35): 7 previously unknown fields were coded and 12 source or procedural details were updated. Other proposed additions, including additions for two further institutions, were deliberately not included. USE Attribution requested: Plagiarism Checker Online, Preliminary research dataset: first 40 institutions, v0.2, 19 Oct 2026, https://www.plagiarism-checker-online.net/university-ai-policy-dataset-preliminary Quoted policy texts remain the property of the institutions.