Use of AI in research and publishing

Generative AI tools are now part of how research is written, reviewed and produced. This policy sets out where we consider their use acceptable, and where it is not. It applies to every journal published under the OAIFR imprint.

AI cannot be an author

A large language model or any other AI tool cannot be listed as an author or co-author. Authorship carries accountability for the work — including responding to questions and correcting the record — and software cannot hold it. Named human authors are fully responsible for everything in the manuscript, including any part produced with AI assistance.

Authors must disclose use

Declare AI use in the manuscript, in the methods where it affected the research and in a statement before the references where it affected the writing. Say which tool, which version, and what it was used for.

Disclosure is expected for:

• Drafting, restructuring or substantially rewriting text.
• Generating or transforming images, figures or diagrams.
• Generating, cleaning or analysing data, and generating code used in the analysis.
• Literature searching or summarising where it shaped the argument.

Disclosure is not required for routine language assistance — spelling, grammar and basic copy-editing tools that do not change the substance.

Author responsibilities

• Verify everything. AI tools produce plausible references that do not exist and confident statements that are wrong. Every citation must be checked against the source. Fabricated references found in a submission are treated as a data integrity problem.
• Do not use AI to generate research data or results and present them as observed.
• Do not use AI to produce images that a reader would take as experimental data, microscopy, imaging or photographic evidence.
• Do not enter confidential or personal data — patient information, unpublished data belonging to others — into a third-party tool.

Reviewers must not upload manuscripts

A manuscript under review is confidential. Uploading it, or any part of it, to a generative AI tool discloses unpublished work to a third party and may put it into a training corpus. This is not permitted under any circumstances.

Reviewers remain personally responsible for the content of their review. A review must reflect the reviewer’s own reading and judgement, and must not be generated by an AI tool.

Editors

Editors must not enter submitted manuscripts, reviewer reports or correspondence into external AI tools. Editorial decisions are made by editors, not by software, and cannot be delegated to a tool.

Detection

We do not treat AI-detection scores as evidence. They are unreliable in both directions, and are particularly prone to flagging authors writing in a second language. Concerns are assessed on the substance of the manuscript — fabricated references, results that do not follow from the methods, or text that does not match the data — through the normal misconduct process.

Changes to this policy

This area is moving quickly and the policy will be revised as practice and guidance develop.