Human responsibility and disclosure

Authors remain responsible for the accuracy, originality and integrity of submitted work. AI tools cannot be listed as authors. Disclose generative AI use that contributed to drafting, analysis, coding, translation, images or other substantive parts of the submission, identifying the tool, version where available, purpose and extent of use. Place research-method uses in Methods and writing assistance in an appropriate declaration; inform the editor in the cover letter. Routine spelling or grammar checks that do not generate substantive content need not be declared.

Human authors must verify generated text, code, calculations and references against the underlying evidence. An AI-generated reference must not be cited without checking that the source exists and supports the claim. AI use does not excuse plagiarism, inaccurate attribution or misleading conclusions. See the ICMJE recommendations on AI use.

AI as a research method

When AI is part of the study, describe the data, model, version or access date, preprocessing, relevant prompts or settings, validation and human oversight sufficiently for assessment. Explain risks of bias, information leakage and limited generalizability. Claims about clinical usefulness require evidence appropriate to the proposed use; development performance alone does not establish patient benefit.

Images, data and confidentiality

Do not use generative AI to fabricate observations, patient records, experimental images or results. Any schematic created with AI assistance must be identified as illustrative, verified by the authors and clearly separated from research evidence. Processing of genuine research images must comply with Image & Data Integrity.

Do not enter identifiable patient information, confidential manuscripts or restricted research data into public AI services without the necessary authorization and protections. Authors must respect consent, access agreements and institutional rules.

Editors and reviewers

Editors and reviewers must not upload submitted manuscripts, unpublished data or review correspondence to external generative AI tools. Scientific judgment and the written assessment remain the responsibility of the assigned person. Any proposed journal-approved assistance must protect confidentiality and be disclosed to the editor. Suspected misuse is assessed under Publication Ethics.