Nature's AI Policy: Disclosure Rules, Image Bans, and What Authors Must Know

Nature requires AI disclosure in the Methods section, not buried in acknowledgments or a footnote. The disclosure should be specific enough that a reader understands the scope of AI involvement. Generic statements like "AI tools were used during writing" aren't sufficient.

A properly formatted disclosure would look something like:

"During preparation of this manuscript, the authors used ChatGPT (OpenAI, GPT-4) to improve the clarity of the Discussion section. After using this tool, the authors reviewed and edited the content as needed and take full responsibility for the final published text."

This placement matters. The Methods section is the part of the paper where other researchers look to understand how work was done. Putting AI disclosure there signals that Nature treats AI use as a methodological choice, not just an administrative detail.

Nature's prohibition on AI-generated images isn't about aesthetics. It's about data integrity. In scientific publishing, images are evidence. A microscopy image, a gel photo, a clinical scan, these are data points that readers use to evaluate claims. AI-generated images can't serve this function because they don't represent real observations.

The ban extends beyond obvious cases. You can't use AI to:

• Generate schematic figures from text descriptions
• Create "enhanced" versions of real images using generative models
• Produce synthetic data visualizations that weren't derived from actual data

AI-assisted image processing tools that enhance existing real images (like denoising or contrast adjustment) occupy a gray area. Nature's guidance suggests these should be disclosed but aren't automatically prohibited, as long as the underlying image data is real and the processing doesn't alter the scientific content.

Nature's AI policy isn't limited to the flagship journal. It applies across the entire Nature Portfolio, which includes:

• All Nature-branded journals (Nature Medicine, Nature Genetics, Nature Methods, Nature Biotechnology, Nature Communications, etc.)
• All Springer Nature publications (Scientific Reports, BMC-series journals, etc.)
• Conference proceedings published through Springer Nature

This means roughly 3,000+ journals follow the same core AI rules. If you've submitted to any Springer Nature journal, you've already been subject to this policy whether you knew it or not.

Nature's AI policy has evolved rapidly:

The speed of this rollout was unusual for academic publishing, where policy changes typically take years. Nature moved within months of ChatGPT's public release to establish rules, and the core principles haven't changed since.

Nature doesn't use AI detection software to screen manuscripts. The editors have publicly acknowledged that current detection tools aren't reliable enough for editorial decisions. Instead, enforcement relies on:

• Author attestation. During submission, authors confirm they've followed the AI disclosure policy. This is a trust-based system, similar to how journals handle conflict of interest disclosures.

• Peer review. Reviewers may flag text that appears AI-generated, though this is informal rather than systematic.

• Post-publication scrutiny. If undisclosed AI use is identified after publication, Nature treats it as a breach of publication ethics, similar to undisclosed conflicts of interest.

The penalty for violating the policy isn't automatic retraction. Nature's editors have indicated they'd handle cases individually, considering the extent of AI use and whether it affected the scientific content. But a violation would damage your relationship with the journal and could trigger a formal investigation under COPE guidelines.

The "yes, but risky" for first draft generation deserves explanation. Nature technically allows it as long as authors take full responsibility for the final text. But in practice, submitting AI-generated drafts with minimal human revision is likely to produce text that reviewers notice, and that undermines confidence in the intellectual contribution of the authors.

In our pre-submission review work with Nature-family manuscripts, we see three specific failure patterns around AI policy compliance.

Disclosure placed in the wrong location. Authors often put AI acknowledgments in the cover letter or acknowledgments when the relevant policy asks for Methods disclosure, or an equivalent manuscript section if no Methods section exists. That makes the disclosure harder for reviewers and readers to interpret as part of the research record.

Copy editing confused with generative drafting. Nature's policy exempts AI-assisted copy editing of human-generated text, but not autonomous content creation or generative editorial work. We see authors under-disclose when they use an LLM to restructure arguments, summarize literature, draft responses, or generate code comments.

Image workflow left undocumented. The visible risk is a fully AI-generated figure, but the quieter risk is using non-generative machine-learning tools to modify real images without caption disclosure. Nature's policy distinguishes generative images from non-generative processing, but both need a careful audit before submission.

Manusights internal analysis shows that the preventable mistake is usually not malicious AI use. It is poor process documentation. Before uploading a Nature Portfolio manuscript, authors should keep a simple record of which tools touched text, code, data analysis, images, and figures.

Before submission:

• Audit your manuscript for any AI-generated content. If you used ChatGPT at any stage, even for brainstorming, decide whether disclosure is warranted.
• Check all figures and images. If any were created or modified using generative AI tools, replace them.
• Draft your Methods section disclosure statement before finalizing the manuscript.

During submission:

• Include the AI disclosure in the Methods section, not in supplementary materials.
• Be specific about which tools you used and which parts of the manuscript they touched.
• If you're unsure whether your use requires disclosure, disclose it anyway. Over-disclosure is always safer than under-disclosure.

Common mistakes to avoid:

• Don't list ChatGPT or another AI tool in the acknowledgments section as if it were a person. Nature's policy specifically addresses this.
• Don't assume that because you heavily edited AI-generated text, it no longer counts as AI-assisted. The use of the tool itself is what requires disclosure.
• Don't use AI to generate supplementary figures thinking the ban only applies to main figures. It applies to all visual content.

Need help preparing your manuscript for Nature's requirements? A Nature submission readiness check can check whether your paper meets the journal's standards 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