If you're writing a research paper in 2026, you've almost certainly used an AI tool at some point — whether for literature search, grammar checking, data analysis, or drafting text. The question isn't whether researchers use AI. It's whether they use it within the rules their target journal has set.
Those rules vary significantly. Some journals are permissive with proper disclosure. Others ban AI-generated images entirely. One thing is universal: every major journal, every ethics body, and every medical editorial committee agrees that AI cannot be an author.
This guide covers the actual policies, not opinions about what the policies should be. We've reviewed the editorial guidelines of every major journal and the positions of ICMJE, COPE, and WAME as of April 2026.
The Universal Rules
Before the journal-by-journal breakdown, here are the points where every major publication agrees:
- AI cannot be listed as an author. Authorship requires accountability, approval of the final manuscript, and the ability to respond to post-publication queries. No AI system can do any of these.
- Human authors bear full responsibility for all content in their manuscript, including anything AI helped produce. If AI introduced a fabricated citation, that's on you.
- Peer reviewers must not upload manuscripts into AI tools. This is treated as a confidentiality breach across all major journals.
- Disclosure is required whenever AI contributed to the substance of the work. The specifics of where and how to disclose vary by journal.
Journal-by-Journal Policies
Nature / Springer Nature
- AI cannot be an author
- Disclosure required in the Methods section — state the tool and how it was used
- Copy editing and grammar checking are exempt from disclosure
- AI-generated images are banned across all Springer Nature journals and books
- Nature does not use AI detection software, acknowledging current tools are unreliable
Science / AAAS
- Among the strictest policies in publishing
- AI use must be noted in the cover letter AND in Methods or Acknowledgments
- Full prompts must be disclosed along with tool name and version
- AI-generated images and multimedia are banned without explicit editor permission
- Undisclosed AI text use is treated as potential misconduct
JAMA / AMA Journals
- AI cannot qualify for authorship
- Disclosure in Acknowledgments, or in Methods if AI was part of research design
- Must include: AI tool name, version, manufacturer, and what content it created or edited
- AI-generated content is discouraged unless part of formal research methodology
- Submission system now asks all authors directly whether AI was used
New England Journal of Medicine (NEJM)
- Disclosure required for AI-assisted technologies used in composing the manuscript
- AI used in clinical research must meet the same evidentiary bar as other interventions
- Clinical trials using AI interventions (enrolled after Jan 1, 2025) must be registered in a WHO-compliant database
- Follows ICMJE authorship guidelines
The Lancet / Elsevier
- AI use disclosed in a separate AI declaration statement
- Basic grammar, spelling, and punctuation checks need no declaration
- The Lancet specifically limits AI to improving readability and language only
- Authors retain full responsibility for all content
IEEE
- AI-generated content permitted with disclosure in Acknowledgments
- Must state: systems used, which sections, level of AI involvement
- Language editing and grammar enhancement exempt from disclosure
- Among the more permissive of the major publishers
PLOS ONE
- AI use declared in Materials & Methods or Acknowledgments
- Must include: tool name, which content it influenced, how outputs were validated
- AI must not be used to fabricate or manipulate data, figures, or images
Quick Reference: Where to Disclose
| Journal | Writing | Images | Disclosure Location | Grammar Exempt? |
|---|---|---|---|---|
| Nature | With disclosure | Banned | Methods | Yes |
| Science | Strict disclosure | Banned* | Cover letter + Methods | No |
| JAMA | Discouraged | With disclosure | Acknowledgments/Methods | Unclear |
| NEJM | With disclosure | With disclosure | Per ICMJE guidelines | Yes |
| Lancet | Language only | With disclosure | AI declaration statement | Yes |
| IEEE | Permitted | With disclosure | Acknowledgments | Yes |
| PLOS | With disclosure | With disclosure | Methods/Acknowledgments | Unclear |
* Science permits AI images only with explicit editor permission, typically limited to papers about AI/ML.
The Three Categories of AI Use
Journals treat different types of AI use very differently. Understanding these distinctions can be the difference between a smooth submission and a desk rejection.
1. Writing and Editing
This is the most common use and the most nuanced. Most journals allow AI writing assistance with disclosure, but the boundaries vary:
- Grammar and spell-checking — Exempt from disclosure at Nature, Elsevier, IEEE, and NEJM. Treat it like any other editing software.
- Language improvement — The Lancet permits AI only for this purpose. If you're a non-native English speaker using AI to polish your prose, most journals consider this acceptable with disclosure.
- Substantive text generation — Having AI draft paragraphs or sections that you then edit requires disclosure everywhere. Science treats undisclosed AI text as potential misconduct. JAMA actively discourages it.
Safe approach: Write your manuscript yourself. Use AI to check grammar, improve clarity, and suggest rephrasing. Disclose if the AI contribution went beyond basic editing. When in doubt, disclose.
2. Data Analysis
AI-assisted data analysis is generally the most accepted use case because it aligns with how researchers already use computational tools. Disclose in your Methods section with:
- The specific tool, version, and date of queries
- The prompts or parameters used
- How you validated the AI's output
- What role the AI played versus your own analysis
This is straightforward because Methods sections already describe analytical tools. AI is just another tool — document it the same way you'd document your statistical software.
3. Image Generation
This is the most restricted category. Springer Nature bans AI-generated images outright. Science bans them without explicit editor permission. The legal uncertainty around AI-generated image copyright is the driving factor — journals don't want to publish images that can't be copyrighted or that may infringe on training data.
Even journals that permit AI images require them to be clearly labeled and disclosed. If you need figures, create them with traditional tools or hire a scientific illustrator. AI-generated figures are not worth the risk at most journals.
What Gets Manuscripts Rejected
Understanding what triggers rejection or retraction helps you avoid it:
- Undisclosed AI use. This is the primary trigger. If reviewers or editors detect AI-generated text that wasn't disclosed, it's treated as a breach of editorial policy — sometimes as misconduct.
- AI hallucinations in citations. AI tools fabricate references. If your paper cites sources that don't exist because an AI generated them, that's grounds for rejection or retraction. Verify every reference.
- Submitting AI-generated images to journals that ban them. Springer Nature and Science will desk-reject for this.
- Telltale AI artifacts. The most infamous case: a paper in Physica Scripta was retracted because the text "Regenerate Response" — copied from the ChatGPT interface — appeared in the manuscript.
- Listing AI as a co-author. Every major journal will reject this immediately.
The Copyright Dimension
The U.S. Copyright Office has ruled that purely AI-generated content cannot be copyrighted. For researchers, this creates a practical concern: if significant portions of your paper were AI-generated, the copyright status of your work is uncertain.
This matters because:
- Journals require copyright transfer or licensing as part of the publication agreement
- If your work contains uncopyrightable AI-generated text, you may not have the rights to transfer
- The more human authorship in the final expression, the stronger the copyright claim
Using AI as a tool in your process — brainstorming, grammar checking, analyzing data — preserves your copyright. Having AI generate the prose weakens or eliminates it.
ICMJE and COPE: The Framework Behind the Rules
Most medical and scientific journals base their policies on guidelines from two bodies:
ICMJE (International Committee of Medical Journal Editors) updated their recommendations in January 2026 with a new dedicated Section V addressing AI use. Their position: AI tools cannot fulfill the four criteria for authorship (substantial contribution, drafting/revision, final approval, accountability). Any AI use in drafting, editing, translation, image generation, or data analysis must be fully disclosed.
COPE (Committee on Publication Ethics) maintains that AI tools cannot take responsibility for submitted work and therefore cannot be authors. They require transparency about AI use in Materials and Methods.
If your target journal's specific AI policy is unclear, following ICMJE and COPE guidelines will keep you safe at virtually any reputable journal.
Practical Checklist for Researchers
Before you submit:
- Read your target journal's specific AI policy. Don't assume — check. Policies differ enough to matter.
- Document your AI use as you work. Note which tools you used, when, what prompts you gave, and what output you used. You'll need this for disclosure.
- Verify every AI-generated reference. Check that cited papers actually exist, are by the stated authors, and say what you claim they say.
- Check for AI artifacts. Search your manuscript for telltale phrases, formatting from chat interfaces, or inconsistencies in technical terminology.
- Prepare your disclosure statement. Include: tool name, version, what it was used for, and what sections it influenced. Place it where your target journal requires.
- Do not list AI as an author or co-author.
- Do not submit AI-generated images to Nature, Science, or any Springer journal.
- Do not use AI tools to generate your peer review if you're serving as a reviewer.
What's always safe:
- Grammar and spell-checking with AI tools
- Using AI to search for and organize literature
- Using AI to help understand complex statistical outputs
- Asking AI to suggest alternative phrasings (then writing your own)
- Using AI-powered reference management tools
What requires disclosure:
- AI-assisted text drafting or substantive editing
- AI-assisted data analysis or interpretation
- AI-generated or AI-modified figures (where permitted)
- AI translation of your manuscript
- Any use of AI that shaped the intellectual content of your paper
How AllScience Fits Into This
If you use AllScience for your research, here is how each of the platform's tools maps onto the disclosure rules above — so you know exactly what you need to say in your cover letter and what you do not.
- Searching for papers: Using AllScience to run one search across seventeen databases is no different from running seven searches by hand. It is literature discovery, not content generation. No journal requires you to disclose which search tool you used.
- Managing citations: Formatting references in APA, Vancouver, Harvard, IEEE, Chicago, MLA, or any of thousands of other styles — this is the same as using Zotero or Mendeley. No disclosure needed anywhere. Every reference in your library is a real paper you saved yourself, with its real DOI and its real authors.
- Checking grammar and style: The 430+ grammar rules that catch passive voice, awkward phrasing, and style drift fall under "copy editing and grammar checking" — which Nature, Elsevier, IEEE, and NEJM all explicitly exempt from disclosure requirements. Same category as running your paper through Grammarly.
- Writing suggestions from the AI: This is the one to be careful with. When AllScience suggests a sentence, it pulls from the papers in your own library and links every claim to its real source — so you never get the hallucinated-citation problem that causes retractions. But if you let the AI draft substantial text, you are in the "AI-assisted text drafting" category and you need to disclose per your target journal's policy. Our advice: use the suggestions the way you would use a colleague's edits. Read them, decide what to keep, rewrite what you keep in your own voice.
The short version: use AllScience the way you would use any other research tool, be honest about where you used it, and you will not run into any of the rejection triggers above.
The Research Tools Most Journals Call "Exempt"
Grammar checking, citation management, literature search, and journal formatting — all the things major journals explicitly exempt from AI disclosure requirements — in one platform.
Start Using AllScienceThis guide reflects policies as of April 2026. Journal policies on AI are evolving — always check your target journal's current guidelines before submission. Links to journal editorial policies are available through AllScience's resource library.