Nature Immunology's AI Policy: What Immunology Authors Need to Know About Disclosure

The Springer Nature AI policy is set centrally, but how it plays out varies by journal because of the research community's characteristics:

The community size point matters for enforcement. Nature Immunology's authorship network is relatively concentrated, the same researchers appear as authors, reviewers, and editorial board members. If your paper has an AI disclosure issue and it becomes known, word travels fast.

Single-cell analysis pipelines

Single-cell immunology is where most AI-related questions arise. A typical Nature Immunology scRNA-seq paper involves:

• Quality control and filtering
• Normalization and batch correction
• Dimensionality reduction (PCA, UMAP)
• Clustering and cell type annotation
• Differential expression analysis
• Trajectory inference
• Gene regulatory network analysis

If you used Copilot or ChatGPT to help write code for any of these steps, that's AI-assisted manuscript preparation (specifically, code generation) requiring disclosure. The analysis tools themselves, Seurat, Scanpy, Harmony, scVI, aren't AI writing tools and don't need to be disclosed under the AI policy (they go in standard Methods as analysis software).

Disclosure example for an scRNA-seq paper:

"Single-cell RNA-seq analysis was performed using Seurat v5 (R) and Scanpy (Python) as described in Methods. GitHub Copilot (Microsoft) was used to assist with writing custom R scripts for the cell type annotation pipeline. All scripts were validated against manual annotation of a reference dataset. ChatGPT (GPT-4, OpenAI) was used to improve the language of the Discussion section. The authors take full responsibility for the published content."

Immune repertoire analysis

TCR and BCR repertoire sequencing generates massive datasets that require specialized computational tools. If AI helped with writing analysis code for clonotype identification, CDR3 analysis, or repertoire diversity calculations, disclose it.

But don't confuse the analysis tools with writing tools. immunarch, MiXCR, and TRUST4 are bioinformatics software described in Methods. ChatGPT helping you write a wrapper script around these tools is AI-assisted code generation requiring separate disclosure.

Clinical immunology and patient data

Nature Immunology publishes translational and clinical immunology research involving patient samples. The same data privacy concerns that apply at Nature Medicine apply here: don't input patient data, clinical records, or identifiable information into cloud-based AI tools.

For papers involving autoimmune disease cohorts, vaccine trial immunogenicity data, or transplant immunology with patient outcomes, keep AI tools away from clinical interpretation sections.

For a systems immunology paper:

"During manuscript preparation, the authors used ChatGPT (GPT-4, OpenAI) to improve the clarity of the Results and Discussion sections. The computational modeling described in this study (immune response ODE model) was developed and parameterized by the authors without AI assistance. The authors take full responsibility for the published content."

For a paper with multi-omics integration:

"The authors used Claude (Claude 3.5, Anthropic) to edit the Introduction for language clarity. GitHub Copilot (Microsoft) assisted with writing Python scripts for integrating CyTOF and scRNA-seq datasets. All integration code was validated against published reference datasets (see Methods: Data Integration Validation). No patient-identifiable data was processed through any cloud-based AI tool."

For a basic immunology paper:

"During preparation of this manuscript, the authors used ChatGPT (GPT-4, OpenAI) to improve the readability of the Methods section. All content was reviewed by the senior author (K.Y.). The authors take full responsibility for the content of the published article."

The flow cytometry gating strategy point is important. Gating decisions in flow and CyTOF experiments are scientific judgments that should be made by trained immunologists. If AI influenced your gating strategy, this isn't just a disclosure issue, it's a scientific integrity question. The analysis code can be AI-assisted, but the gating logic should come from the investigators.

Standard Springer Nature enforcement:

• During review: Request to add disclosure; possible rejection if concealment appears deliberate
• After publication: Correction, expression of concern, or retraction depending on severity

Immunology community dynamics: Nature Immunology publishes approximately 150-200 research articles per year. The immunology community is small enough that most researchers in a given subfield know each other. A publication ethics issue at this journal has outsized reputational consequences compared to a megajournal.

If your paper involves collaboration between immunology labs and computational biology groups, which is increasingly common, make sure both sides contribute to the AI disclosure. The wet-lab PI may not know that the bioinformatics collaborator used Copilot for the analysis pipeline, and the bioinformatics lead may not know that the first author used ChatGPT to rewrite the Introduction.

Immunity (Cell Press) uses STAR Methods format, so your AI disclosure goes in Method Details rather than a free-form Methods section. The Journal of Immunology (AAI) follows Oxford University Press guidelines. Frontiers in Immunology follows Frontiers' publisher-wide policy. The substantive requirements are similar across all five journals, the differences are in formatting and enforcement culture.

For authors choosing between Nature Immunology and Immunity: the AI policies are functionally equivalent. The disclosure format differs (Methods vs. STAR Methods), but the rules and consequences are the same.

Nature Immunology adopted Springer Nature's AI policy when it was formalized in early 2023. The core principles haven't changed since:

This stability matters for immunology researchers planning multi-year projects. If you're midway through a study that won't be submitted for another year, the current rules are very likely the rules you'll face at submission time. Science (AAAS) went through multiple policy revisions, initially banning all AI text, then switching to a disclosure model. Springer Nature's approach was always disclosure-based from the start, so Nature Immunology authors haven't had to deal with moving goalposts.

For single-cell immunology papers:

• Keep a record of which analysis code was AI-assisted and which was written from scratch
• If your cell type annotations used any AI-based automatic annotation tools, describe these as analysis methods, not in the AI writing disclosure
• Validate all AI-generated code against published reference datasets before including in the paper

For vaccine and clinical immunology:

• Don't process participant immunological data through cloud AI tools
• Keep AI away from efficacy claims and clinical outcome descriptions
• If your paper includes immune correlates of protection, the interpretation should be entirely human-driven

For mechanistic immunology papers:

• AI can help with language, but descriptions of signaling pathways, knockout phenotypes, and molecular mechanisms should reflect what you observed, not what an LLM predicts based on training data
• Be careful with AI-edited Discussion sections, LLMs may insert claims about immune mechanisms that sound plausible but don't match your specific experimental system

Before submission checklist:

• [ ] AI disclosure in Methods with tool names, versions, use cases
• [ ] Research tools (Seurat, Scanpy, FlowJo) described in standard Methods
• [ ] Writing/code tools (ChatGPT, Copilot) in separate AI disclosure
• [ ] No patient/participant data processed through cloud AI
• [ ] No generative AI images or pathway diagrams
• [ ] All co-authors aware of AI disclosure
• [ ] AI-assisted code validated independently

A free manuscript assessment can help verify your Nature Immunology submission meets editorial and disclosure standards.