AI-generated image policies of 100 high-impact journals and their publishers, 2025–2026 (dataset)

This dataset documents how the 100 journals with the highest two-year mean citedness in OpenAlex, and the publishers whose policies they inherit, regulate AI-generated and AI-altered images. Journal author instructions and cited publisher AI/image policies were captured on 30 September 2026. Archived baselines from September–October 2025 were retrieved from the Internet Archive, and policy change windows in 2026 were dated from archived captures. Rules are coded for five image types (explanatory illustrations, data visualisations, primary research images, graphical abstracts/table-of-contents graphics, and cover art) using six stances (prohibited, restricted, permitted with conditions, permitted with disclosure, general AI rule only, not addressed). Every code is linked to a verbatim quotation from the source document and was checked against the captured text by an automated substring test. Files: journal sampling frame (frame.csv), coded journal-level policies (policies.csv), publisher-level codes (publishers.json), journal-level overrides and inconsistency flags (journal_overrides.json), capture log and source URLs (sources.csv, journal_sources.csv), Internet Archive captures used (wayback_captures.csv), codebook (codebook.md) and the capture/analysis scripts (code/). Full-text page captures are not redistributed for copyright reasons and are available from the author on request.