PNAS AI Policy: National Academy Rules for America's Broadest Science Journal
PNAS requires AI disclosure in both Methods and Author Contributions, prohibits AI authorship, and applies the same rules across all submission tracks including the NAS contributed track.
Senior Researcher, Oncology & Cell Biology
Author context
Specializes in manuscript preparation and peer review strategy for oncology and cell biology, with deep experience evaluating submissions to Nature Medicine, JCO, Cancer Cell, and Cell-family journals.
Next step
Choose the next useful decision step first.
Use the guide or checklist that matches this page's intent before you ask for a manuscript-level diagnostic.
PNAS occupies a distinctive spot in scientific publishing. It's the journal of the National Academy of Sciences, an institution whose members include over 300 Nobel laureates. It publishes across every scientific discipline. And it still has a "contributed" submission track where NAS members can communicate papers directly. This institutional weight means PNAS's AI policy isn't just another publisher guideline, it carries the imprimatur of America's most prestigious scientific body. The rules are straightforward, but they have an unusual quirk: PNAS is one of the few major journals that requires AI disclosure in two separate locations.
The PNAS AI policy
PNAS sets its own AI policy under the authority of the NAS. The core rules:
- AI can't be an author. Consistent with ICMJE guidelines, AI tools don't meet authorship criteria. They can't take accountability, approve manuscripts, or be responsible for scientific claims.
- AI use must be disclosed in Methods AND Author Contributions. This dual requirement distinguishes PNAS from most competitors. The Methods section provides the detailed description; the Author Contributions statement flags AI involvement for readers scanning the paper.
- AI-generated images are prohibited. No figures or visual content produced by generative AI tools.
- Authors bear full responsibility. Every listed author must vouch for all content, including AI-assisted sections.
- The policy applies to all submission tracks. Direct submission, contributed track, and presubmission inquiry papers all follow the same rules.
The dual-disclosure requirement
Most journals require AI disclosure in one place, usually Methods. PNAS requires it in two:
Methods section (detailed):
"During the preparation of this manuscript, the authors used ChatGPT (GPT-4, OpenAI) to improve the clarity and readability of the Introduction and Discussion sections. GitHub Copilot (Microsoft) was used to assist with writing Python scripts for the phylogenetic analysis. All AI-suggested text was reviewed by the authors, and all code was validated against published reference datasets. The authors take full responsibility for the published content."
Author Contributions statement:
"A.B. designed the study. C.D. performed experiments. E.F. analyzed data. A.B. and C.D. wrote the manuscript with language editing assistance from ChatGPT (GPT-4, OpenAI)."
Why two places? PNAS's Author Contributions section is a prominent feature of every paper, it's one of the first things readers check. Including AI disclosure there ensures visibility. The Methods section provides the detail. Together, they give readers both the flag and the specifics.
This dual requirement also means that incomplete disclosure is easier to detect. If your Methods section mentions AI use but your Author Contributions doesn't (or vice versa), an editor or reviewer will notice the inconsistency.
How the NAS institutional context matters
PNAS isn't published by a commercial publisher. It's published by the National Academy of Sciences, which advises the federal government on science policy. This institutional identity affects AI policy in several ways:
Credibility stakes. When PNAS publishes a paper, it carries the NAS's institutional credibility. Undisclosed AI use that's later discovered doesn't just embarrass the authors, it reflects on the Academy.
The contributed track. NAS members can "contribute" papers to PNAS through a streamlined review process. The AI policy applies equally to contributed papers, but the contributed track has historically faced scrutiny for potentially lighter review. Adding undisclosed AI to a contributed paper amplifies existing concerns about that track's rigor.
Policy influence. NAS positions on science policy carry weight in Washington. If PNAS's AI policy were seen as lax, it could undermine the Academy's credibility in advising on AI governance in science. The incentive to enforce the policy firmly is institutional, not just editorial.
Writing the disclosure for PNAS
For a biological sciences paper:
Methods:
"During manuscript preparation, the authors used ChatGPT (GPT-4, OpenAI) to improve the language of the Results and Discussion sections. All AI-suggested text was reviewed by the corresponding author (G.H.) and verified against the experimental data. No AI tools were used for data analysis or experimental design."
Author Contributions:
"G.H. and I.J. designed the study. K.L. performed experiments. M.N. analyzed data. G.H. and K.L. wrote the manuscript with language editing assistance from ChatGPT."
For a physical sciences paper:
Methods:
"The authors used Claude (Claude 3.5, Anthropic) to edit the Introduction for language clarity. Density functional theory calculations were performed using Quantum ESPRESSO 7.2 without AI assistance."
Author Contributions:
"O.P. designed and performed calculations. Q.R. supervised the project. O.P. and Q.R. wrote the manuscript with language editing from Claude."
For a social sciences paper:
Methods:
"During preparation of this manuscript, the authors used ChatGPT (GPT-4, OpenAI) to assist with improving the readability of the statistical methods description. The survey instrument, data collection, and statistical analysis were conducted without AI assistance."
Author Contributions:
"S.T. designed the survey. U.V. collected data. W.X. performed statistical analysis. All authors wrote the manuscript with language assistance from ChatGPT."
What requires disclosure at PNAS
Use case | Disclosure required? | PNAS-specific notes |
|---|---|---|
Grammar/spell check | No | Standard tools exempt |
ChatGPT for language editing | Yes | Methods + Author Contributions |
AI for data analysis code | Yes | Both locations; confirm validation |
AI as research subject | No (research method) | Standard Methods |
AI-generated figures | Prohibited | Data-derived plots fine |
Translation of manuscript | Yes | Name tool and languages |
AI for supplementary text | Yes | Part of the submission |
AI for significance statement | Yes | PNAS requires a 120-word significance statement |
AI for cover page brief description | Gray area | Describe if substantial AI involvement |
AI for statistical analysis | Yes | Specify which analyses |
The significance statement point is PNAS-specific. Every PNAS paper includes a brief "Significance" statement that explains the paper's importance to a general scientific audience. If AI helped write this, which is tempting given the 120-word constraint, disclose it. This statement is prominently displayed and is often the first thing readers see.
Consequences of non-disclosure
PNAS follows COPE guidelines with NAS institutional weight:
During review:
- Editor contacts corresponding author
- Disclosure must be added to both Methods and Author Contributions
- Inconsistency between the two disclosure locations raises additional concerns
- Deliberate concealment can result in rejection
After publication:
- Correction for minor undisclosed language editing
- Expression of concern for unclear scope
- Retraction for fabricated content or false claims
- NAS institutional review for serious cases involving Academy members
The contributed track risk: If a contributed paper is found to have undisclosed AI use, the scrutiny extends to both the authors and the NAS member who contributed it. The member vouched for the paper's quality, and an integrity issue reflects on their judgment.
Cross-disciplinary visibility: PNAS publishes across all sciences. A retraction or correction is visible not just to specialists but to the entire scientific community. The journal's broad readership means AI-related issues at PNAS tend to generate more attention than similar issues at a specialty journal.
Comparison with other broad-scope journals
Feature | PNAS | Science | Nature | Science Advances | Nature Communications |
|---|---|---|---|---|---|
Publisher | NAS | AAAS | Springer Nature | AAAS | Springer Nature |
AI authorship | Prohibited | Prohibited | Prohibited | Prohibited | Prohibited |
Disclosure location | Methods + Author Contributions | Acknowledgments/Methods | Methods | Acknowledgments/Methods | Methods |
Dual disclosure | Yes | No | No | No | No |
AI image ban | Yes | Yes | Yes | Yes | Yes |
Former AI text ban | No | Yes (Jan–Nov 2023) | No | Yes (Jan–Nov 2023) | No |
Contributed/member track | Yes | No | No | No | No |
PNAS is the only journal in this comparison with a dual-disclosure requirement and a member-contributed track. The combination creates a unique accountability structure that's tighter than most competitors.
How the policy compares across PNAS submission tracks
Track | AI disclosure required? | Review process | Additional notes |
|---|---|---|---|
Direct submission | Yes (Methods + Author Contributions) | Standard peer review | Most submissions |
Contributed track | Yes (Methods + Author Contributions) | NAS member-arranged review | Member accountability applies |
Presubmission inquiry | N/A (no manuscript yet) | Editorial assessment | AI policy applies at full submission |
The contributed track deserves emphasis. If you're an NAS member contributing a paper, your AI disclosure obligations are identical to direct submissions. The track doesn't provide any exemptions or relaxed requirements. Given ongoing debates about the contributed track's future, any integrity issue in a contributed paper attracts disproportionate attention.
Practical advice for PNAS submissions
For all submissions:
- Prepare two disclosure statements: a detailed one for Methods and a brief one for Author Contributions
- Make sure they're consistent, if Methods mentions ChatGPT, Author Contributions should too
- Don't forget the Significance statement. If AI helped write it, say so.
For biological sciences papers:
- If AI helped with genomics, proteomics, or metabolomics code, disclose in both locations
- Deposit code in a public repository, PNAS expects code availability
For physical sciences and engineering:
- Simulation code written with AI assistance needs disclosure
- Theoretical derivations should be human-generated; AI can edit the text but shouldn't produce the math
For social sciences:
- If AI helped with statistical code (R, Stata, Python), disclose and confirm validation
- Survey instruments and qualitative analysis should be human-designed
For contributed track papers:
- Ensure the contributing NAS member is aware of all AI use in the manuscript
- The member's reputation is on the line alongside yours
Before submission checklist:
- [ ] AI disclosure in Methods section
- [ ] AI disclosure in Author Contributions statement
- [ ] Disclosures are consistent between both locations
- [ ] Tool name, version, and use case specified
- [ ] No AI-generated images
- [ ] Significance statement AI use disclosed if applicable
- [ ] Code deposited in public repository
- [ ] All co-authors aware of AI disclosure
A free manuscript assessment can help verify your PNAS submission meets the journal's dual-disclosure and formatting requirements.
Sources
Reference library
Use the core publishing datasets alongside this guide
This article answers one part of the publishing decision. The reference library covers the recurring questions that usually come next: how selective journals are, how long review takes, and what the submission requirements look like across journals.
Dataset / reference guide
Peer Review Timelines by Journal
Reference-grade journal timeline data that authors, labs, and writing centers can cite when discussing realistic review timing.
Dataset / benchmark
Biomedical Journal Acceptance Rates
A field-organized acceptance-rate guide that works as a neutral benchmark when authors are deciding how selective to target.
Reference table
Journal Submission Specs
A high-utility submission table covering word limits, figure caps, reference limits, and formatting expectations.
Before you upload
Choose the next useful decision step first.
Move from this article into the next decision-support step. The scan works best once the journal and submission plan are clearer.
Use the scan once the manuscript and target journal are concrete enough to evaluate.
Anthropic Privacy Partner. Zero-retention manuscript processing.
Where to go next
Conversion step
Choose the next useful decision step first.
Use the scan once the manuscript and target journal are concrete enough to evaluate.