The Lancet's AI Policy: Stricter Than Elsevier, With an Anti-AI-Imagery Stance
The Lancet lets authors use AI tools with human oversight, asks for a separate AI declaration at the end of the manuscript, limits AI-made figures, and bans AI in World Report and Perspectives pieces.
Readiness scan
AI policy covered? See what The Lancet reviewers are likely to push back on.
The AI policy governs disclosure, not how the science is judged. Run the Free Readiness Scan with The Lancet as your target to see desk-rejection risk, likely reviewer concerns, and journal fit.
Quick answer: The Lancet allows AI tools in manuscript preparation as long as a human stays in control, and it wants that use declared in a separate AI declaration statement at the end of the manuscript, submitted with the paper and printed with it. Name the tool and version, what you used it for, and how you checked its output. Basic grammar, spelling and punctuation checks need no declaration (The Lancet information for authors, August 2026 edition, checked October 8, 2026).
The stricter Lancet rules sit in its opinion sections. World Report and Perspectives pieces may not use AI at all, and Comments and Correspondence may use it only for grammar and spelling.
The Lancet AI Policy at a Glance
- AI authorship: Not allowed. AI tools cannot be listed or cited as authors.
- Where to declare: A separate AI declaration statement at the end of the manuscript, made at submission. AI used in the research itself is described in the Methods.
- What to declare: Tool name and version, purpose, and the extent of your oversight. Editors may ask for your prompts.
- No declaration needed: Basic grammar, spelling and punctuation checks.
- Figures: Generative AI mainly for brainstorming image concepts; no generated images of real people or copyrighted material.
- Opinion pieces: No AI in World Report or Perspectives; grammar and spelling only in Comments and Correspondence.
The policy structure
The Lancet's current rules for research papers come down to five points:
1. AI may support the work, not replace the thinking. The instructions say AI tools "must never be used as a substitute for human critical thinking, expertise, and evaluation." You are accountable for anything a tool produced, which includes checking its references, because AI-generated citations can be wrong or invented.
2. Declare it in its own statement. The declaration goes at the end of the manuscript, is submitted with the paper, and is published with it. It should give the tool's name and version, the purpose, and how far you oversaw the output, with prompts where applicable.
3. Research use goes in the Methods. If AI was part of how the study was done (an analysis model, a classifier, a screening tool), describe it in detail there. Trials of AI interventions also follow CONSORT-AI, and their protocols SPIRIT-AI.
4. AI can't be an author. Authorship carries responsibilities only humans can hold.
5. Be careful with images and confidential material. No generated images that copy copyrighted work or show real people, and check a tool's terms before you paste in an unpublished manuscript or identifiable data.
A Lancet Global Health letter on AI imagery
In March 2026, Lancet Global Health published a Correspondence titled "Beyond disclosure: stop using AI imagery in global health," by Imogen Bakelmun and Kent Buse. It responds to a November 2025 piece by Arsenii Alenichev and colleagues on what they call "poverty porn 2.0," the spread of AI-generated depictions of human suffering across the global health industry. It is a letter by its authors, who declare competing interests, not an editorial and not a statement of Lancet policy.
The letter's specific concerns:
- AI imagery perpetuates a regressive visual lexicon of racialised tropes
- It undermines consent and agency for the people the images claim to portray
- It erases the labour and livelihoods of local image makers
The authors say disclosing the use of AI is "necessary, but insufficient," and propose a practical standard for global health communications: no use of AI imagery depicting people or contexts of vulnerability, commissioning local photographers from the communities depicted, and adopting dignity-first, enforceable image policies. It is a strong anti-AI-imagery position, published in a Lancet journal, but it does not change the rules in The Lancet's information for authors.
How The Lancet's policy differs from Elsevier's general rules
For a research article, less than older summaries suggest. The AI section of The Lancet's instructions opens with "Authors preparing a manuscript for an Elsevier journal can use AI tools to support them," and the core requirements (human oversight, accountability, a declaration, no AI authors) follow that publisher framework. The Lancet-specific layer is in two places:
Aspect | Research articles | Lancet opinion sections |
|---|---|---|
AI use | Allowed with human oversight | None in World Report or Perspectives; grammar and spelling only in Comments and Correspondence |
Declaration | Separate AI statement at the end of the manuscript | Required at submission if any AI was used |
Confidential data | Check the tool's terms first | Never paste unpublished data from a linked article; a breach means rejection |
Source: The Lancet information for authors, August 2026 edition, checked October 8, 2026.
The practical upshot: a correspondence letter is the place where a Lancet submission is most likely to fall foul of the AI rules, because translation, shortening and rewording by AI are all ruled out there.
Research integrity at The Lancet Group
The Lancet Group's research integrity page says its commitment is led internally by a Deputy Editor of The Lancet who is also the research integrity and risk management lead, supported by a working group of editors and technical experts who monitor cases and external developments. When an article is submitted, editors check for statements on patient consent and conflicts of interest and use tools to screen for plagiarism and simultaneous submissions.
The page does not describe a separate AI-specific integrity body. For a journal that publishes some of the most influential medical research in the world (The Lancet's Journal Impact Factor is 109), maintaining an integrity function makes sense. The stakes of a retraction or misconduct finding at The Lancet are career-defining.
What this means for different manuscript types
The Lancet publishes several article types, and the AI policy implications vary:
Original research articles. AI help is allowed with oversight and a declaration. The safe test is whether you could defend every sentence, number and reference without the tool open.
Reviews and meta-analyses. If AI screened or extracted studies, that is research use and belongs in the Methods, not only in the end-of-manuscript declaration.
Comments and correspondence. Grammar and spelling help only. The instructions bar AI for translating the piece, improving arguments, generating sentences, reordering, shortening or summarising, and a declaration is required if any AI was used. World Report and Perspectives allow none.
Scope across Lancet journals
The policy applies to all journals in the Lancet family:
- The Lancet (IF ~109)
- Lancet Oncology (IF ~33.7)
- Lancet Infectious Diseases (IF ~29.4)
- Lancet Neurology (IF ~54.6)
- Lancet Psychiatry (IF ~21.1)
- Lancet Global Health (IF ~22.5)
- Lancet Digital Health (IF ~25.5)
- Lancet Respiratory Medicine (IF ~34.7)
- Lancet Public Health
- Lancet HIV
- Lancet Haematology
- Lancet Gastroenterology & Hepatology
- Lancet Child & Adolescent Health
- Lancet Healthy Longevity
- Lancet Microbe
- Lancet Rheumatology
- Lancet Regional Health (all editions)
- eBioMedicine
- EClinicalMedicine
That's a substantial portfolio. If you're submitting to any Lancet-branded journal, these rules apply regardless of the specific title's subject area or impact factor.
Practical disclosure guidance
Put the statement at the end of the manuscript under its own heading. A statement that covers what The Lancet asks for:
"AI declaration: During preparation of this manuscript the authors used ChatGPT (OpenAI, GPT-4o) to improve the readability of the Introduction and Discussion. The authors reviewed and edited all output, checked every reference against the original source, and take full responsibility for the content."
Things to include:
- The tool name and version
- The purpose, section by section if the use was uneven
- The extent of your oversight, and prompts if the editors ask
Things to avoid:
- Vague descriptions ("AI tools were used")
- Folding the statement into the acknowledgments or the cover letter instead of the separate declaration
- Declaring grammar-checker use you do not need to declare, while leaving out a model that drafted text
Readiness check
Run the scan while the topic is in front of you.
See score, top issues, and journal-fit signals before you submit.
The Lancet vs. other top medical journals
Feature | The Lancet | NEJM | BMJ | JAMA |
|---|---|---|---|---|
AI scope | Assistance with human oversight; none in World Report or Perspectives | Writing assistance with disclosure | Writing assistance with disclosure | Writing assistance with disclosure |
Disclosure location | Separate AI declaration at end of manuscript | Cover letter + manuscript | Methods or acknowledgments | Cover letter + manuscript |
AI images | Brainstorming use; no copied or real-person likeness | Not explicitly addressed in detail | Banned | Not explicitly addressed in detail |
Dedicated integrity group | Yes (a research integrity lead and working group) | No (uses broader publication-ethics guidance) | No (uses broader publication-ethics guidance) | No (uses broader publication-ethics guidance) |
Parent publisher policy | Elsevier framework plus opinion-section limits | Independent (NEJM Group) | Stricter than BMJ general | Independent (AMA) |
The Lancet column reflects its August 2026 instructions; the other columns are summaries from earlier checks, so confirm them on each journal's own page before you rely on them.
Advice for Lancet submissions
Write the declaration before you upload, not at the form. Go back through the drafting history and list every tool that touched text, analysis or figures. A grammar checker can stay off the list; a model that redrafted the Discussion cannot.
Check every AI-suggested reference. The Lancet names fabricated references as an author responsibility. One citation that does not exist is enough to make an editor doubt the rest.
Keep confidential material out of public tools. Read the tool's terms on training and data use before pasting an unpublished manuscript or patient-level data into it.
Treat correspondence as the strict case. If you are writing a Comment or a letter, do the translation, shortening and argument yourself.
Submitting to a Lancet journal? A The Lancet submission readiness check can help ensure your manuscript meets editorial standards before you enter the formal review process.
Bottom line
For a Lancet research paper, AI help is allowed when a human checks it and the use is declared in a separate statement at the end of the manuscript. The hard limits are in the opinion sections: no AI in World Report or Perspectives, and only grammar and spelling help in Comments and Correspondence. Older summaries that say "acknowledgments only" or "readability only" describe a previous version of the instructions.
What should you do about The Lancet's AI policy?
Comply proactively if:
- You used any AI tool (ChatGPT, Grammarly, Copilot) during manuscript preparation
- The journal requires AI use disclosure (The Lancet asks for a separate statement at the end of the manuscript)
- 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
The Lancet at a glance
The figures The Lancet publishes about speed, selectivity and cost, each linked to the page we read it on. They describe the journal, not your manuscript's chances.
Frequently asked questions
Yes, with human oversight. The Lancet's information for authors (August 2026 edition) says AI tools can support authors but must never substitute for human critical thinking, expertise and evaluation, and authors stay accountable for every AI-assisted output, including checking references that AI may have fabricated.
In a separate AI declaration statement at the end of the manuscript, made at submission; it appears in the published article. Name the tool and version, the purpose and the extent of your oversight, and supply prompts where applicable. AI used in the research itself goes in the Methods. Basic grammar, spelling and punctuation checks need no declaration.
The Lancet says generative AI should be used mainly for brainstorming image concepts, with direct use in figures limited to demonstrating what the tool can do, and images that copy copyrighted material or depict real people must not be generated. Separately, Lancet Global Health published a March 2026 Correspondence by two outside authors urging global health communicators to stop using AI imagery that depicts people or contexts of vulnerability; it is the authors' proposal, not a Lancet policy.
Yes. The policy applies across the Lancet family: The Lancet, Lancet Oncology, Lancet Infectious Diseases, Lancet Neurology, Lancet Psychiatry, Lancet Global Health, Lancet Digital Health, Lancet Respiratory Medicine, and all other Lancet titles.
The Lancet's general rules follow Elsevier's framework: AI may assist with human oversight and must be declared. Where The Lancet goes further is its opinion sections. AI is not allowed at all in World Report and Perspectives, and Comments and Correspondence may use it only for grammar and spelling, with a declaration.
Sources
Free guide
AI cannot be an author. None of the 49 policies we checked allows it.
Not a co-author, not an acknowledged contributor with a byline, not "with assistance from". What they do require is a disclosure, and the wording matters. The guide has the statement filled in, with what each publisher asks for beside it.
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.
Private API processing. Your manuscript is not used to train models.
Put the guidance to work
Choose the practical next step for your research.
Start with the question in front of you, or use the guided learning path when you want a clearer view of the full publishing process.