Journal AI policy tracker: publisher and journal rules on AI authorship, disclosure, images and peer review, edition 2026-09

A tracker of 49 AI policies issued by publishers, imprints, societies and journals, and the 100 journals they govern. For each policy it records whether an AI tool may be listed as an author; whether authors must disclose AI use; whether AI-generated images or figures are allowed; whether peer reviewers may use AI tools; whether peer reviewers may upload a manuscript to an AI tool; whether editors may use AI tools, with a paraphrase of the rule, the page that states it and the date it was read. Checked through 2026-10-25. Method: Each policy row was checked against the publisher's, society's or journal's own web page, fetched and read on the verified_on date. Values are paraphrased; see each field's source URL. "not_stated" means the page(s) read do not address the point; "unverified" means no official page could be read, usually because it returned 403 or a bot challenge, which was not bypassed. Third-party compilations were used only to locate official URLs. Journal rows map each journal to the policy that governs it; mapping_basis says whether that link was confirmed from the policy page or inferred from the publisher of record (Crossref). Journal-specific instruction pages were not read unless named as the policy_ref, so overrides are not recorded in this edition. Sources marked retrieved_via "browser_read" were read on their verified_on date in a standard browser after the automated fetch was refused. Human-verification challenges were never completed; a value is recorded only where the policy page itself was read. Scope: Twenty publishers: large scholarly publishers by journal output together with the major medical, general-science and computing publishers named in the F2 brief (size was not re-ranked for this edition); the imprints, societies and journals whose own statements govern the journals below; and 100 journals ranked by Manusights search demand: every journal with a Manusights AI-policy post, plus the journals with the most Search Console impressions across Manusights pages (90 days to 2026-06-03). Counts: 49 of 49 policy pages read; 238 of 294 policy cells sourced, 56 not addressed by the pages read, 0 unverified. 100 of 100 journals map to a read policy. Values paraphrase each policy; they are not the issuers' text, and each links its source. Policies change: check the linked page before relying on a rule, and cite it when you quote one. Journals marked publisher_of_record were matched to their publisher's policy and may add their own rules. Files: one row per journal as CSV, every policy, field, source and note as JSON, a codebook, README, licence, CITATION.cff and SHA-256 checksums. At release, the CSV and JSON are byte-identical to the downloads on the canonical page. Canonical page (updated between versions): https://manusights.com/blog/journal-ai-policies-2026