Artificial intelligence policy

Launch preparation · Editorial proposals dated 25 October 2026. Public submissions are closed. Policies and article limits await scientific and publisher ratification.

Human responsibility

AI systems cannot be authors, accountable editors or independent peer reviewers. Human authors remain responsible for every statement, result, reference, image and interpretation. Assistance must not create fabricated data, invented references, false participant records or misleading images. AI output is not a primary source of scientific evidence.

What authors must declare

At submission, state whether AI assisted the work. For relevant use, give the provider, tool/model, version or access date, purpose, affected parts and human verification. Explain generative language editing in the cover letter and acknowledgements. Describe assistance in programming, data analysis or scientific methods in Methods, with sufficient detail for reproducibility, including validated code, relevant settings and prompts where appropriate and lawful.

• Language: distinguish basic spelling checks from generative rewriting or translation; declare generative assistance, verify meaning and protect confidential content.
• Programming: identify generated code, tests, human review and the final software version; authors must understand and validate the analysis.
• Data analysis: describe the model, inputs, validation, performance, uncertainty and bias checks. Synthetic data used as a legitimate research method must be labelled and explained, never passed off as observed patient data.
• Scientific content: disclose generated passages, summaries or other substantive content; verify the source literature and do not delegate interpretation. Generated clinical or experimental imagery must not be presented as observed evidence.

Reviewers and editors

Unpublished manuscripts are confidential. Do not upload them, their figures, reviews, identifiers or participant information to external AI systems without the journal's prior approval, adequate confidentiality guarantees and any necessary author permission. Human editors retain decision authority. Any authorised editorial use will be documented; authors will be informed about substantive AI assistance in editing or production. Automated detection alone cannot establish misconduct.

Standard declaration

During preparation of this work, the authors used [provider, tool/model, version or date] for [purpose], affecting [sections, code or analyses]. The authors reviewed and verified the outputs by [checks], including all cited sources and relevant calculations. [Describe confidentiality safeguards and any methodological details.] The authors take full responsibility for the submitted work.

If no relevant tool was used: “The authors declare that no generative AI was used in the preparation or analysis of this work.” Declare basic non-generative checking separately if it is material to the methods. Update the statement on revision. Undeclared or inappropriate use will be assessed under the journal's integrity procedures, proportionately to the facts.

Prepared against ICMJE author guidance, ICMJE editorial guidance and COPE's position on AI and authorship; checked 25 October 2026.