JCI's AI Policy: What Authors Need to Know About AI Disclosure at the Journal of Clinical Investigation

Most journals covered in AI policy guides inherit their rules from a parent publisher. Nature Medicine follows Springer Nature. Cancer Cell follows Cell Press. Gut follows BMJ Publishing Group. The JCI is different, it's an ASCI journal with editorial independence.

What this means in practice:

No publisher-wide cascade. When Springer Nature updates its AI policy, all 3,000+ journals follow automatically. When the JCI updates its policy, it's a decision by the JCI's editorial board. This makes changes slower but more deliberate.

ICMJE alignment is the foundation. The JCI has long followed ICMJE recommendations closely. Its AI policy builds directly on the ICMJE's position on AI and authorship, rather than being filtered through a corporate publishing policy.

The JCI Insight connection. JCI Insight is the JCI's open-access companion journal. Both journals follow the same AI policy set by the ASCI editorial board. If you've read the JCI's rules, you've read JCI Insight's rules.

The JCI publishes research that bridges basic science and clinical medicine, the "bench to bedside" work that physician-scientists specialize in. This creates specific AI considerations:

JCI papers frequently include human subject data: patient cohorts, biomarker analyses, clinical correlations, genetic studies. The same data privacy rules that apply at clinical journals apply here: don't input patient data into cloud-based AI tools.

For translational papers that include both mouse model data and human patient data, the AI disclosure should address both:

"No human participant data or animal experimental records were processed through cloud-based AI tools. ChatGPT (GPT-4, OpenAI) was used to improve the language of the Discussion section. All AI-suggested text was reviewed by the clinical investigators (J.R. and S.M.)."

The JCI's editorial identity centers on mechanistic insight with disease relevance. Papers don't just describe phenomena, they connect molecular mechanisms to clinical conditions. If AI tools helped write sections that make these connections, the disclosure should reflect this, and the claims themselves should be entirely human-generated.

AI-polished prose about a signaling pathway is fine if the scientific content is accurate. AI-generated claims about how that pathway connects to a disease mechanism aren't acceptable without careful human verification.

Many JCI papers integrate genomics, transcriptomics, proteomics, and metabolomics data. If AI helped with analysis code for any of these, disclose it:

"The authors used GitHub Copilot (Microsoft) to assist with writing R scripts for the RNA-seq differential expression analysis and the GSEA pathway enrichment analysis. All scripts were validated against published tutorial datasets and manually verified by the bioinformatics team (L.K.). The multi-omics integration pipeline was developed without AI assistance."

For a translational research paper:

"During preparation of this manuscript, the authors used ChatGPT (GPT-4, OpenAI) to improve the readability of the Introduction and Discussion sections. No AI tools were used for data analysis, clinical interpretation, or mechanistic conclusions. All AI-suggested text was reviewed by the corresponding author (A.B.) and verified against the experimental data. The authors take full responsibility for the published content."

For a clinical investigation paper:

"The authors used Claude (Claude 3.5, Anthropic) to edit the Methods section for language clarity. No patient data was processed through cloud-based AI tools. The statistical analysis was performed using SAS 9.4 by the study biostatistician (C.D.). Clinical outcome assessments and their interpretations are entirely the work of the investigator team. The authors take full responsibility for the content."

For a basic science paper with disease relevance:

"During manuscript preparation, ChatGPT (GPT-4, OpenAI) was used to assist with editing the Results section. GitHub Copilot (Microsoft) helped write Python scripts for the ChIP-seq peak calling analysis. All code was validated against ENCODE reference datasets. The disease mechanism described in the Discussion is based on the experimental data and was formulated by the authors without AI assistance."

The JCI follows COPE and ICMJE guidelines for handling publication ethics issues:

During review:

• Editor contacts corresponding author
• Disclosure must be added to Methods
• Deliberate concealment may result in rejection
• If AI use affected clinical or mechanistic claims, additional review may be required

After publication:

• Correction for minor undisclosed language editing
• Expression of concern if scope is unclear or if it affects scientific content
• Retraction for fabricated data or false mechanistic/clinical claims
• Institutional notification for serious cases

The physician-scientist community: The ASCI has approximately 3,400 active members, and many serve as JCI reviewers and editorial board members. The community is small enough that publication ethics issues become known quickly. A correction at the JCI affects not just the paper but the author's standing within the ASCI membership, a community that elects its members based on scientific achievement and integrity.

Science Translational Medicine (AAAS) follows the same AI policy as Science. PNAS has its own policy set by the National Academy of Sciences, which requires AI disclosure in both Methods and Author Contributions. The JCI's approach, Methods-only disclosure with ICMJE alignment, is the most straightforward of the group.

For translational papers:

• Clearly separate AI writing tools from research computational tools
• If your paper connects a molecular mechanism to a disease, the mechanistic reasoning must be entirely human-generated
• Disclose AI use in the cover letter as a courtesy, even though the JCI primarily requires Methods disclosure

For papers with patient cohorts:

• Don't process patient data through cloud AI
• If AI helped with analysis code for clinical data, confirm that the code was validated by the study biostatistician
• Clinical phenotyping and outcome assessments should be described by the clinical team, not AI

For JCI Insight submissions:

• Same policy as the JCI. Don't look for different rules.
• JCI Insight is open access, so your disclosure is publicly visible to all readers

Before submission checklist:

• [ ] AI disclosure in Methods with tool name, version, use case
• [ ] No patient data processed through cloud AI tools
• [ ] Clinical and mechanistic interpretations are human-generated
• [ ] Analysis code validated independently
• [ ] All co-authors aware of AI disclosure
• [ ] ICMJE authorship criteria confirmed for all listed authors

A JCI submission readiness check can help verify your JCI submission meets the journal's editorial and ethical requirements before you submit.

Comply proactively if:

• You used any AI tool (ChatGPT, Grammarly, Copilot) during manuscript preparation
• The journal requires AI use disclosure in the methods or acknowledgments
• Your institution has its own AI use policy that may be stricter

Less concerned if:

• You used AI only for grammar/spell checking (most journals exempt this)
• The journal does not have a formal AI policy yet
• Your use was limited to literature search or reference management