Taylor & Francis Lets AI Draw a Diagram but Not Touch a Result

Taylor & Francis’s AI policy separates four questions that other publishers sometimes blur together: whether AI can be an author, what has to be said about using it to write, what it may touch in a figure, and what a reviewer or editor may do with it.

Authorship. “Generative AI tools must not be listed as an author”, because authorship carries “uniquely human responsibilities”: accountability, consent to publication, and contractual assurances about the work’s integrity that a tool cannot give.

Disclosure. “Authors must clearly acknowledge within the article or book any use of Generative AI tools”, and the acknowledgement has to name the tool’s full name and version number, how it was used, and why. A journal article needs a specific AI usage disclosure statement; a book author tells their editorial contact about planned AI use at the earliest stage, and if approved, the statement goes in the preface or introduction rather than being added later.

Figures. This is where the policy draws its clearest line. AI assistance is permitted for research data visualisations, conceptual illustrations of processes, flow diagrams, and teaching illustrations. The policy “does not permit the use of Generative AI in the creation or manipulation of outputs of research”, and defines manipulation broadly — adding, hiding, moving, removing or introducing features in an image or figure.

Reviewers and editors. Reviewers must not upload unpublished manuscripts or project proposals, including associated files, images or information, into Generative AI tools, and must not use AI to generate review reports — though they may use it to improve their own review language, and remain fully responsible for the review’s accuracy. Editors carry the same no-upload rule, and should check with their Taylor & Francis contact before using any Generative AI tool not already authorised.

The figures rule is the one worth sitting with, because it is a genuinely useful distinction rather than a blanket ban. A diagram explaining a process is an argument about how something works, and AI assistance there is a drafting aid like any other. A microscopy image or a clinical sample photograph is a claim about what was observed, and altering it with AI would be altering the evidence itself — which is why the policy treats the two completely differently rather than regulating “AI in images” as one category.

Naming the tool and version in the disclosure, rather than accepting a generic “AI was used”, is what makes the declaration checkable later: tool versions change their behaviour and their error patterns, and a vague disclosure cannot be investigated if a question comes up after publication.

The book-specific timing rule — disclose to the editorial contact at the earliest stage, not at submission of the final manuscript — reflects how differently books are produced: a book contract is agreed long before the manuscript exists, so the point to settle AI use is before writing starts, not after it is finished.

How it reaches a specific journal. The AI policy is published at group level and applies across Taylor & Francis’s journals and book imprints; individual journals implement the disclosure through their own submission systems, typically a dedicated AI-use field alongside the cover letter.

Where a journal is society-owned, the society may add further conditions; it cannot remove the baseline rules set out here.