Taylor & Francis (part of Informa) issued expanded guidance on generative AI for authors, editors, and reviewers on 12 June 2024, consolidating and clarifying rules that had previously been spread across individual journal instructions. The policy sits inside the publisher’s broader AI Policy and applies across its journals portfolio (Taylor & Francis, Routledge, and imprint journals published on tandfonline.com) as well as its books program. This guide sets out, in operational terms, what the policy actually requires of an author submitting to a Taylor & Francis title: whether an AI tool can be named as an author, exactly what must be disclosed and where, and how the requirement differs for books versus journal articles.
Can a generative AI tool be listed as an author?
No. Taylor & Francis’s position is that generative AI tools cannot meet the responsibilities that come with authorship: they cannot take accountability for the submitted content, cannot manage copyright and licensing agreements, and cannot give the contractual assurances a publishing agreement requires. This tracks the same underlying reasoning used across the industry — ICMJE, COPE, and every major publisher CASRAI has reviewed converge on the same conclusion for the same reason: authorship carries accountability obligations a software tool structurally cannot hold. See CASRAI’s ICMJE generative AI policy entry for the source position most publisher policies, including this one, trace back to.
Practically, this means: an AI tool (ChatGPT, Claude, Gemini, or any other generative system) cannot appear in the author byline, cannot be assigned corresponding-author responsibilities, and cannot be a signatory on the publishing agreement. The human author who used the tool remains fully responsible for the accuracy, originality, and integrity of everything in the submission, including content the AI tool helped produce.
What must be disclosed, and where
Where an author has used a generative AI tool in preparing a journal article, Taylor & Francis requires a disclosure statement that names three things:
- The specific tool used, including its version number (e.g. “ChatGPT-4” rather than just “ChatGPT”, or “Grammarly’s generative AI feature” rather than just “Grammarly”) — a generic reference to “AI assistance” without naming the tool does not satisfy the requirement.
- How the tool was used — e.g. to draft specific passages, to suggest edits to language and phrasing, to generate or refine code, to summarize source material.
- The reason for using it — the purpose the tool served in producing the submission.
For journal article submissions, this statement belongs in the Methods section where the journal has one, or in the Acknowledgments section where it does not. It should not be folded into the author-contribution statement in place of, or confused with, a generative AI disclosure statement more generally, which CASRAI’s dictionary entry covers as a cross-publisher concept. Authors should also check the individual journal’s own instructions for authors, since some Taylor & Francis journals layer additional field-specific detail on top of the group-wide baseline.
Book authors: earlier and more proactive disclosure
The requirement for book authors is structured differently from the journal-article requirement, reflecting the longer production timeline and the closer, more ongoing editorial relationship in books publishing. Book authors intending to use generative AI tools must disclose that intent to their editorial contact at the earliest possible point — at the proposal stage if the intended use is already known then, or during the manuscript-writing stage if the intention forms later — and must obtain approval before proceeding. This is a disclose-and-seek-approval model rather than the disclose-in-the-published-text model used for journal articles, though the published disclosure obligation (naming the tool, how, and why) still applies once the manuscript is submitted.
Where the policy draws the line: prohibited uses
Beyond the authorship and disclosure rules, Taylor & Francis’s guidance restricts specific uses of AI tools regardless of disclosure. Authors, editors, and reviewers are directed not to use generative AI in ways that could compromise the accuracy of the scholarly record, breach confidentiality (for example, entering unpublished manuscript text or peer-review content into a public AI tool), infringe proprietary rights, or violate data protection obligations. This last point matters in practice for anyone reviewing a Taylor & Francis submission: pasting a confidential manuscript into a consumer-facing AI chatbot for a quick summary is treated as a confidentiality breach independent of whatever the disclosure rules say about authorship.
Across the whole policy, the constant is author accountability: whatever role an AI tool played, the named human authors remain accountable for the originality, validity, and integrity of the content submitted, including the accuracy of references and citations — a generative AI tool’s tendency to fabricate plausible-looking citations does not shift responsibility for catching that error away from the author.
How this compares to other major publishers
Taylor & Francis’s core position — no AI authorship, mandatory named-tool disclosure, author retains full accountability — is broadly consistent with the position taken by Elsevier, Springer Nature, Wiley, PLOS, the Royal Society of Chemistry, the American Chemical Society, and Cell Press, all of which CASRAI surveys in Journal and Publisher Policies on Generative AI in Manuscripts: The Policy Landscape. The detail that varies most across publishers, and the detail worth checking journal-by-journal before submission, is exactly where the disclosure statement must appear (Methods vs. Acknowledgments vs. a dedicated AI-use declaration field in the submission system) and how granular the “how it was used” description needs to be. Taylor & Francis’s Methods-or-Acknowledgments placement rule and its version-number naming requirement are specific enough that a disclosure written for one publisher’s template will often need editing, not just copy-pasting, before it satisfies another’s.
Practical checklist for authors submitting to a Taylor & Francis journal
- Confirm no AI tool is listed as an author or corresponding author.
- Write a disclosure statement naming the specific tool and version, describing how it was used, and stating why.
- Place that statement in the Methods section if the journal has one; otherwise in Acknowledgments.
- Check the specific journal’s Instructions for Authors for any field-specific additions to the group-wide policy.
- For book proposals or manuscripts, contact the editorial team before using generative AI tools, not after.
- Verify every AI-assisted citation and reference against the primary source before submission — the disclosure requirement does not reduce the author’s responsibility for accuracy.
- Do not paste confidential manuscript or peer-review text into a public-facing AI tool.
Frequently asked questions
Does Taylor & Francis allow AI-generated images or figures?
The group-wide AI policy’s authorship and disclosure requirements apply to AI-generated or AI-assisted text and content broadly; image and figure integrity is also covered under Taylor & Francis’s separate image-manipulation and research-integrity policies, and any AI-generated visual content should be disclosed using the same tool-name-and-purpose approach described above. Authors working with AI-generated images should check the specific journal’s figure-integrity guidance in addition to the group AI policy.
What happens if AI use is not disclosed?
Undisclosed AI use that surfaces after publication is handled the same way any other undisclosed material fact or research-integrity concern is handled — through Taylor & Francis’s standard post-publication investigation process, which can range from a correction to a retraction depending on severity, consistent with COPE’s guidelines on handling such cases.
Does this policy apply to peer reviewers as well as authors?
Yes. Reviewers and editors are subject to the same confidentiality-driven restrictions — principally, not entering unpublished manuscript content into external AI tools — even though the authorship and named-disclosure rules are specifically framed around authors.
Is this the same as the ICMJE position on AI and authorship?
The underlying reasoning is the same (AI tools cannot hold accountability, so cannot be authors), but the specific disclosure mechanics differ by publisher. See CASRAI’s ICMJE generative AI policy entry and the cross-publisher policy landscape guide for how the mechanics vary.
