AI use and editorial integrity

Authors must be able to explain and defend the work they submit. AI assistance does not transfer that responsibility to a tool.

Policy version 1 November 2026. Applies to new submissions and updated declarations at revision. Editors will clarify requirements for manuscripts already under consideration.

What authors must declare

At submission, state whether AI was used in preparing the manuscript or any accompanying material. Include the cover letter, figures, supplements, code, and point-by-point response. Report the tool and provider, model or version where available, access dates, purpose, affected sections or documents, and the human checks performed. Update the declaration when use changes at revision.

Use a dedicated AI assistance declaration in the manuscript and a matching summary in the cover letter. Describe AI used for research design, analysis, coding or model evaluation in Methods with enough detail to assess reproducibility. Distinguish the AI studied in the research from AI used to prepare the submission. Ordinary non-generative spelling correction alone may be declared as spelling correction only.

Permitted assistance and author responsibility

Language editing, translation, organization, coding assistance and other support can be considered when transparently reported and checked by qualified authors. Authors must verify every claim, citation, numerical result and interpretation against the original sources and analyses. AI systems cannot be authors. A disclosure does not make fabricated results, invented citations, plagiarism, misleading images or unsupported conclusions acceptable.

Substantial drafting must be described explicitly. The responsible authors must demonstrate the intellectual contribution, verify the underlying work and meet authorship requirements. Do not submit unchecked generated prose or use rewriting tools to conceal assistance. Do not generate or alter research images in a way that misrepresents evidence. Any AI-based image processing that is part of the research method requires transparent methods and preservation of originals.

Confidentiality

Do not place identifiable patient information, confidential manuscripts, review reports or private editorial correspondence into public or unapproved AI tools. Editors and reviewers must obtain journal authorization before processing confidential material with AI, use only approved private systems, declare substantive assistance, and retain responsibility for their own assessment.

Screening and fair assessment

JAIM may use approved tools to support integrity screening. Detector outputs are signals for editorial review; they do not establish authorship or misconduct. JAIM does not apply a percentage cutoff for rejection. Short texts, boilerplate, language editing and scientific terminology can affect results. Tools and versions must be recorded, and unsupported inputs should be left unscored.

An editor examines the underlying work and any specific concerns, gives authors a chance to explain, and records the evidence and decision. Requests for supporting methods, code, sources or revisions must relate to the concern. Authors may appeal to an unconflicted senior editor through the editorial office. Decisions rest on scientific quality, accuracy, transparency and accountable human contribution.

Declaration format

If no relevant AI was used, state: “No AI-assisted technology was used to prepare this submission or its accompanying documents.” If only ordinary non-generative spellcheck was used, state that explicitly. Otherwise identify the tools, dates, purpose, affected components and human verification. Use the AI disclosure form or preparation tool.

Policy basis

This JAIM policy is informed by ICMJE guidance and Nature Portfolio policy. JAIM sets its own requirements. Published research has documented false positives in evaluated detectors; see Liang and colleagues. That study does not measure the performance of every current detector.