NEJM's AI Policy: Dual Disclosure, ICMJE Alignment, and the NEJM AI Twist

NEJM's AI policy is grounded in the ICMJE recommendations, which most major medical journals follow. The ICMJE issued guidance in 2023 stating that AI tools don't meet authorship criteria and that their use should be disclosed. NEJM adopted this framework directly.

The ICMJE authorship criteria require that each author:

1. Made substantial contributions to conception/design or acquisition/analysis/interpretation of data
2. Drafted the work or revised it critically for important intellectual content
3. Approved the final version
4. Agrees to be accountable for all aspects of the work

AI tools can arguably contribute to criterion 2 (drafting), but they can't meet criteria 3 or 4. An LLM can't approve a final version with understanding, and it certainly can't be held accountable if problems emerge post-publication. That's why every ICMJE-aligned journal, including NEJM, excludes AI from authorship.

NEJM's two-location requirement is more than Nature's one (Methods section) but less than Science's three (cover letter, acknowledgments, methods). Here's what it looks like:

In the cover letter:

"We wish to disclose that during preparation of this manuscript, we used Claude (Anthropic) for editing the Discussion section to improve clarity. The authors reviewed and verified all AI-edited content and take full responsibility for the manuscript."

In the manuscript (typically in the acknowledgments or a disclosure section):

"The authors used Claude (Anthropic) for language editing during manuscript preparation. All content was reviewed and verified by the authors."

The cover letter disclosure is especially important because NEJM's editors use it to assess the submission before sending it out for review. If they discover undisclosed AI use later, it raises questions about what else might not have been disclosed.

Here's where things get interesting. In 2024, the NEJM Group launched NEJM AI, a journal dedicated to artificial intelligence in medicine. NEJM AI's editorial policy on AI use in manuscripts is strikingly different from the flagship journal's cautious approach.

NEJM AI explicitly encourages authors to use LLMs during manuscript preparation. The journal's editors have stated they welcome experimentation with AI tools and consider their use a natural extension of the computational methods NEJM AI publishes about. Authors must still disclose AI use, but the tone is permissive rather than regulatory.

This creates a useful contrast within the same publishing group:

The difference makes sense. NEJM AI's audience consists of researchers who build and study AI systems. Telling them not to use the tools they work on would be absurd. The flagship NEJM's audience is broader: clinicians, epidemiologists, and clinical trialists who may or may not have AI expertise.

NEJM doesn't publicize specific penalties for AI policy violations. The approach appears to be case-by-case, with the editorial team assessing the severity and intent of any undisclosed AI use. This is more moderate than Science's blanket "misconduct" classification.

Enforcement mechanisms include:

• Cover letter screening. Editors note AI disclosures during initial manuscript review.
• Peer reviewer flags. Reviewers who suspect undisclosed AI use can raise concerns during review.
• Post-publication review. If issues emerge after publication, NEJM can issue corrections or, in severe cases, retract papers.

NEJM's editorial team has signaled that they're more concerned about AI use that affects scientific content (fabricated data, incorrect clinical recommendations) than about AI used for language polishing. The severity of response likely scales with the impact on the science.

NEJM and JAMA share similar positions: both allow general AI writing assistance, require dual disclosure (cover letter + manuscript), and follow ICMJE guidelines. The Lancet and Cell Press are more restrictive, limiting AI to readability only.

Before submission:

• Keep a record of any AI tool use during manuscript preparation. You'll need specifics for both disclosure points.
• Review all AI-assisted text carefully. NEJM's editors evaluate manuscripts at the highest level of scientific rigor. AI-generated text that contains factual errors or clinical inaccuracies will be caught during review.

The cover letter:

• Mention AI use prominently, don't bury it at the bottom. Editors appreciate transparency.
• Be specific about which sections of the manuscript were AI-assisted.
• State clearly that all authors reviewed and take responsibility for the content.

The manuscript disclosure:

• Place it in the acknowledgments or a designated disclosure section. Check the current NEJM author guidelines for the preferred location, as this can change.
• Match the disclosure to what you wrote in the cover letter. Inconsistencies between the two documents would raise editorial concerns.

What to avoid:

• Don't use AI to generate clinical recommendations or treatment conclusions. For a journal whose audience includes practicing physicians, AI-generated clinical content carries patient safety implications that go beyond publication ethics.
• Don't assume NEJM AI's permissive stance applies to the flagship journal. They're different publications with different policies.
• Don't skip the cover letter disclosure thinking the manuscript disclosure is sufficient. NEJM specifically requires both.

For non-native English speakers:

• AI language polishing is fully supported. The dual disclosure requirement applies, but there's no stigma attached to using AI for English improvement. Editors understand that language barriers shouldn't prevent good science from being published.

Submitting a clinical manuscript to NEJM? A free manuscript review can help you verify that your paper meets the journal's standards, including proper AI disclosure formatting.

NEJM's moderate AI policy reflects the journal's institutional character: evidence-based, measured, and focused on accountability rather than prohibition. By grounding its policy in the ICMJE framework, NEJM ensures alignment with virtually every major medical journal that follows the same guidelines. And by launching NEJM AI with an AI-friendly editorial stance, the NEJM Group acknowledges that AI tools are becoming integral to medical research while maintaining appropriate guardrails for clinical content in the flagship journal.

The practical takeaway is straightforward: use AI tools if they help your manuscript, disclose in two places, don't list AI as an author, and make sure every claim in your paper can withstand scrutiny from reviewers who expect the highest standard of evidence in medicine.