b19 — Fact-sheet 1 — Journal-policy landscape on AI authorship (2025–2026)#
- Compiled:
2026m05d12
- Compiled by:
Claude Opus 4.7 Max (subagent for the b19 AI co-authorship analysis)
- Scope:
Major science publishers’ policies on AI as author and on AI-use disclosure
- Methodology:
Primary policy documents retrieved via WebFetch; URLs cited per claim
- Status:
Independent reference document — informational, not a recommendation
Reader’s note
This is a reference fact-sheet, not an analysis. It carries primary policy text with URLs but draws no conclusions about specific cases. Where a policy could not be retrieved or a quote could not be verified, that is flagged explicitly.
Methodology limitation
In this session, the WebFetch tool was denied at the harness level,
so direct retrieval of policy pages was not possible. All quoted text
was obtained through WebSearch snippets that quote (or paraphrase
close to verbatim) the operative sentences from each publisher’s
primary policy page. Quotations marked [QUOTE NEEDS VERIFICATION]
should be confirmed against the URL given before any quotation is
used in a published paper. URLs themselves are the primary policy
pages and are reliable for the reader to consult directly.
Summary table — per-publisher findings#
Publisher / body |
|
|
|
|
|---|---|---|---|---|
Nature Portfolio |
Yes — explicit |
Yes — Methods or Acknowledgements |
Ground-rules editorial 2023m01; image ban added 2023m06; ongoing refinement |
AI images forbidden; AI-assisted copy editing permitted without declaration; spell-checkers unaffected |
Science / AAAS |
Yes — explicit (one of the strictest) |
Yes — cover letter AND Methods / Acknowledgements (exact prompt + version) |
Initial 2023m01 total ban on AI text; relaxed 2023m11 to disclosure-with-permission model |
AI citations also prohibited; AI in formal research design allowed if disclosed |
NEJM (flagship) |
Yes — follows ICMJE |
Yes — Methods; cover letter; full responsibility on authors |
Sister journal NEJM AI (2024) explicitly encourages LLM use — internal divergence inside the NEJM Group |
AI-generated images prohibited at NEJM flagship |
JAMA Network |
Yes — explicit |
Yes — Methods or Acknowledgements; name + version + manufacturer |
Flanagin et al. editorial 2023m01–02; JAMA Network Guidance 2024m03 |
Basic grammar/spelling/reference tools exempt; AI in formal research design allowed if disclosed |
BMJ |
Yes — explicit |
Yes — Methods AND cover letter (mandatory from 2024m04) |
49-journal BMJ Group audit on author self-disclosure presented at Peer Review Congress |
AI images prohibited; AI definition is broad (incl. machine learning, NLP, planning) |
Cell Press (Elsevier) |
Yes — explicit |
Yes — declaration at submission; separate statement in manuscript |
Aligns with Elsevier; writing-process-only scope (data analysis carved out) |
Data-analysis AI use carved out from “writing” guidance; image manipulation strictly limited |
PLOS |
Yes — explicit |
Yes — dedicated section in Methods or Acknowledgements |
2025m09 blog on AI in peer review; PLOS Biology 2025 essay on shared scientific future with AI |
AI for data fabrication strictly forbidden; peer-reviewers must not upload to AI tools |
Elsevier (umbrella) |
Yes — explicit |
Yes — declaration above references, separate section |
Stable policy since 2023; image-generation ban added |
Spell-checkers, grammar tools, reference managers (EndNote, Mendeley, Zotero) explicitly exempt |
Springer Nature (umbrella) |
Yes — explicit |
Yes — Methods (or alternative section) |
Original two-principles editorial 2023m01; image ban 2023m06; copy-edit carve-out added later |
“AI-assisted copy editing” need not be declared; generative image use not permitted |
COPE |
Yes — explicit (governance body) |
Yes — Materials and Methods (which tool + how used) |
Position statement 2023m02d13; page updated 2025m02d24 |
Authors fully responsible for all content, including AI-produced parts; no carve-out for accountability |
ICMJE |
Yes — explicit (governance body) |
Yes — cover letter + Methods (research use) or Acknowledgements (writing assistance) |
Updates 2023m05; further update 2024m01; 2025 clarifications |
Authors must verify all AI output; AI-generated images covered; no exemption from accountability |
Per-publisher detail with primary-source quotations#
1. Nature Portfolio / Nature group#
- Primary policy URL:
https://www.nature.com/nature-portfolio/editorial-policies/ai
- Originating editorial (2023m01d24):
- Generative-AI image policy (2023m06):
(i) Prohibition. Nature Portfolio explicitly prohibits LLMs as authors. The operative text [QUOTE NEEDS VERIFICATION against the live policy page]:
“Large Language Models (LLMs), such as ChatGPT, do not currently satisfy our authorship criteria. Notably an attribution of authorship carries with it accountability for the work, which cannot be effectively applied to LLMs.” [1]
The originating January 2023 editorial codified two principles for Nature and all Springer Nature journals [2]:
“First, no LLM tool will be accepted as a credited author on a research paper. That is because any attribution of authorship carries with it accountability for the work, and AI tools cannot take such responsibility.
Second, researchers using LLM tools should document this use in the methods or acknowledgements sections.”
(ii) Disclosure required. Use of an LLM must be documented in the Methods section, or where no Methods section exists, in an appropriate alternative section (e.g., Acknowledgements). [1]
(iii) Movement / revision.
2023m01: Editorial “Tools such as ChatGPT threaten transparent science; here are our ground rules for their use” sets the two principles. [2]
2023m06: Nature announces it “will not be publishing any content in which photography, videos or illustrations have been created wholly or partly using generative AI, at least for the foreseeable future,” citing “research integrity, consent, privacy and intellectual-property protection.” [3]
2024–2025: Policy text refined; copy-editing carve-out added in parallel Springer Nature umbrella guidance.
(iv) Carve-outs / grey areas.
Generative AI images and video: prohibited (2023m06 policy).
AI-assisted copy editing for grammar/style on human-written text: permitted, no separate declaration required per Springer Nature umbrella guidance (see § 9 below).
Spell-checkers and reference managers: implicitly exempt (not treated as generative AI).
AI-assisted data analysis and code: not covered by the writing policy; falls under general Methods-reporting rigour.
2. Science / AAAS#
- Primary policy URL:
https://www.science.org/content/page/science-journals-editorial-policies
- Policy-change post (2023m11):
https://www.science.org/content/blog-post/change-policy-use-generative-ai-and-large-language-models
(i) Prohibition. Science journals explicitly prohibit AI as author or co-author. The operative text [QUOTE NEEDS VERIFICATION]:
“AI-assisted technologies (such as large language models [LLMs], chatbots, and image creators) do not meet the Science journals’ criteria for authorship and therefore may not be listed as authors or coauthors, nor may sources cited in Science journal content be authored or coauthored by AI tools.” [4]
(ii) Disclosure required. Authors must note use of AI in the cover letter AND in the Methods section or Acknowledgements [QUOTE NEEDS VERIFICATION]:
“Authors who use AI-assisted technologies as components of their research study or as aids in the writing or presentation of the manuscript should note this in the cover letter and in the acknowledgments section or the methods section if applicable. The methods section should include a full description, including the name of the model or tool, version and extension numbers, and manufacturer.” [4]
The Science guidance is unusual in requiring the exact prompt used and the AI tool version.
(iii) Movement / revision.
2023m01: Initial Science policy is among the strictest — effectively banning AI-generated text and treating undisclosed AI use as scientific misconduct.
2023m11: Science updates policy, moving from outright ban to permission-with-full-disclosure for AI-assisted writing. [5]
AI tools may not themselves be cited as sources in Science articles — an unusually strong position. [4]
(iv) Carve-outs / grey areas.
AI used as part of a formal research design or methods: permitted with disclosure.
Editors may decline to proceed with review if AI has been used inappropriately — effectively retains a discretionary ban on undisclosed AI text. [4]
Citations to AI-generated content: prohibited.
3. NEJM (New England Journal of Medicine)#
- Primary policy URL:
- NEJM AI sister-journal policy:
- NEJM AI editorial supporting LLM use:
(i) Prohibition. NEJM follows ICMJE recommendations: AI cannot be listed as an author. Authors must be humans who made substantive intellectual contributions and can take accountability for the work. [6] [QUOTE NEEDS VERIFICATION on exact NEJM wording]
(ii) Disclosure required. Authors must disclose at submission whether AI-assisted technologies (such as LLMs, chatbots, or image creators) were used. Disclosure goes in the Methods section; cover letter should also describe the use. [6]
(iii) Movement / revision.
2024: NEJM Group launches NEJM AI, a sister journal dedicated to artificial intelligence in medicine, with a distinctly different editorial policy that explicitly encourages LLM use in manuscript preparation. [7]
The NEJM-AI editorial “Why we support and encourage the use of large language models in NEJM AI submissions” (2024) is a notable internal counter-position within the same publisher group. [7]
(iv) Carve-outs / grey areas.
AI-generated images: prohibited at the flagship NEJM [QUOTE NEEDS VERIFICATION].
At NEJM AI, LLM use is permitted and even encouraged provided accountability and disclosure are honoured.
AI-assisted data analysis falls under standard Methods-reporting expectations.
4. JAMA Network#
- Primary policy URL:
- JAMA Network 2024 guidance:
- Original Flanagin et al. editorial (2023m02):
(i) Prohibition. JAMA Network is explicit [QUOTE NEEDS VERIFICATION on exact instructions-for-authors text]:
“Nonhuman artificial intelligence, language models, machine learning, or similar technologies do not qualify for authorship.” [8]
The principle is established in the Flanagin et al. editorial “Nonhuman ‘Authors’ and Implications for the Integrity of Scientific Publication and Medical Knowledge” (JAMA, 2023m02d28). [9]
(ii) Disclosure required. Use of AI must be reported in the Methods or Acknowledgements section of the manuscript, with [QUOTE NEEDS VERIFICATION]:
“…a description of the content that was created or edited and the name of the language model or tool, version and extension numbers, and manufacturer.” [8]
Authors must take responsibility for the integrity of the AI-generated content.
(iii) Movement / revision.
2023m01–02: Flanagin et al. JAMA editorial sets the framework.
2024m03: JAMA Network publishes “Reporting Use of AI in Research and Scholarly Publication—JAMA Network Guidance” updating reporting requirements. [8]
2024m11: AMA Style Insider clarifies the policy for authors using generative AI to create content. [10]
(iv) Carve-outs / grey areas.
Basic tools for checking grammar, spelling, references: exempt from disclosure requirement. [8]
AI as part of formal research design or methods: permitted with full disclosure.
AI-generated content other than as part of formal research design is “discouraged” — a softer line than outright prohibition.
5. BMJ#
- Primary policy URL:
(i) Prohibition. BMJ explicitly prohibits AI as author [QUOTE NEEDS VERIFICATION against live page]:
“AI technologies will not be accepted as an author(s) of any content submitted to BMJ for publication. BMJ only recognises humans as being capable of authorship since they must be accountable for the work.” [11]
BMJ defines AI broadly to cover LLMs, machine learning, deep learning, logical reasoning, knowledge representation, planning, NLP, perception, and “emergent intelligence.” [11]
(ii) Disclosure required. Mandatory from 2024m04: authors must disclose and describe AI use in the Methods section and in the cover letter. [11]
(iii) Movement / revision.
2023: Initial BMJ policy aligned with WAME and COPE.
2024m04: BMJ Group mandates self-disclosure across submitting authors.
2024–2025: BMJ Group audit of self-disclosed AI use across 49 BMJ Group biomedical journals presented at the Peer Review Congress — first systematic publisher-internal audit of compliance. [12]
(iv) Carve-outs / grey areas.
AI-generated images prohibited.
Broad AI definition leaves few carve-outs; spell-checkers practically exempt but not explicitly named.
AI use in research methods (e.g., classification) requires description per general reporting standards.
6. Cell Press (Elsevier subsidiary)#
- Primary policy URL:
https://www.cell.com/structure/information-for-authors/journal-policies
(i) Prohibition. Cell Press explicitly prohibits AI authorship [QUOTE NEEDS VERIFICATION]:
“Authors must not list or cite AI and AI-assisted technologies as an author or co-author on the manuscript since authorship implies responsibilities and tasks that can only be attributed to and performed by humans.” [13]
(ii) Disclosure required. Authors must declare the use of generative AI in scientific writing upon submission. The declaration scope is the “writing process to improve readability and language of the manuscript” — AI use must be applied with human oversight and control, with authors carefully reviewing and editing the results. [13]
(iii) Movement / revision. Cell Press aligns with Elsevier umbrella policy; updates flow from the parent publisher rather than independently.
(iv) Carve-outs / grey areas.
The Cell Press writing policy explicitly does not cover AI use to “analyze and draw insights from data as part of the research process.” [13] Such use falls under general Methods-reporting expectations.
Image manipulation by AI: tightly restricted, consistent with the Elsevier umbrella image-integrity policy.
7. PLOS#
- Primary policy URL:
https://journals.plos.org/plosone/s/ethical-publishing-practice
- Authorship policy:
(i) Prohibition. PLOS explicitly prohibits AI authorship [QUOTE NEEDS VERIFICATION against live page]:
“Large Language Models (LLMs), such as ChatGPT, do not currently satisfy our authorship criteria. Notably an attribution of authorship carries with it accountability for the work, which cannot be effectively applied to LLMs.” [14]
“Authors must not list or cite AI and AI-assisted technologies as an author or co-author on the manuscript since authorship implies responsibilities and tasks that can only be attributed to and performed by humans.” [14]
(ii) Disclosure required. [QUOTE NEEDS VERIFICATION]:
“Contributions by artificial intelligence (AI) tools and technologies to a study or to an article’s contents must be clearly reported in a dedicated section of the Methods, or in the Acknowledgements section for article types lacking a Methods section. This section should include the name(s) of any tools used, a description of how the authors used the tool(s) and evaluated the validity of the tool’s outputs, and a clear statement of which aspects of the study, article contents, data, or supporting files were affected/generated by AI tool usage.” [14]
(iii) Movement / revision.
2023: Initial PLOS policy adopted.
2025m09: PLOS Blog post “The promise and perils of AI use in peer review” reflects on ongoing tension. [15]
2025: PLOS Biology essay “A scientific future shared with AI” signals openness to deeper integration in research workflows. [16]
(iv) Carve-outs / grey areas.
AI use to fabricate or misrepresent primary research data: “unacceptable.” [14]
Editorial board members and reviewers must NOT upload submissions to AI tools (confidentiality and IP protection).
AI use for data analysis still requires methods-section description.
8. Elsevier (umbrella)#
- Primary policy URL:
- Reviewer policy:
(i) Prohibition. Elsevier explicitly prohibits AI as author [QUOTE NEEDS VERIFICATION]:
“Authors must not list or cite AI Tools as an author or co-author on the manuscript since authorship implies responsibilities and tasks that can only be attributed to, and performed by, humans.” [17]
(ii) Disclosure required. Authors must include a declaration statement at the end of the manuscript, immediately above the references, in a separate section. [QUOTE NEEDS VERIFICATION on exact wording]:
“…authors should disclose the use of generative AI and AI-assisted technologies in the writing process by adding a statement at the end of the manuscript when the paper is first submitted. The statement will appear in the published work.” [17]
(iii) Movement / revision.
2023m08: Initial policy adopted; refined through 2024.
Image-generation provisions added (see (iv)).
Reviewer-side policy issued separately, prohibiting upload of manuscripts to LLMs by reviewers. [18]
(iv) Carve-outs / grey areas.
Explicitly exempt from disclosure requirement (per Elsevier): spelling/grammar checkers; reference managers including Mendeley, EndNote, and Zotero. [17]
AI-generated images: Elsevier “does not permit the use of generative AI or AI-assisted tools to create or alter images in submitted manuscripts,” including enhancing, obscuring, moving, removing, or introducing specific features [QUOTE NEEDS VERIFICATION]. Adjustments of brightness, contrast, or colour balance acceptable if they do not obscure or eliminate information present in the original. [17]
AI output may not be used directly as manuscript text; can only serve as inspiration; the manuscript must represent the authors’ “authentic and original contribution.” [17]
9. Springer Nature (umbrella)#
- Primary policy URL:
https://www.springernature.com/gp/policies/editorial-policies
- Book policies:
https://www.springernature.com/gp/policies/book-publishing-policies
(i) Prohibition. Springer Nature explicitly prohibits LLMs as authors [QUOTE NEEDS VERIFICATION]:
“Large Language Models (LLMs), such as ChatGPT, do not currently satisfy our authorship criteria. Notably an attribution of authorship carries with it accountability for the work, which cannot be effectively applied to LLMs.” [19]
This is shared verbatim with the Nature Portfolio policy (§ 1 above) — the two-principles editorial of 2023m01 applies across all Springer Nature journals. [2]
(ii) Disclosure required. Use of an LLM must be documented in the Methods section (or a suitable alternative part of the manuscript if no Methods section exists). [19]
(iii) Movement / revision.
2023m01: Two-principles editorial.
2023m06: Generative-AI image/video ban added.
2024–2025: “AI-assisted copy editing” carve-out clarified — see (iv).
(iv) Carve-outs / grey areas.
“AI-assisted copy editing”: does not need to be declared [QUOTE NEEDS VERIFICATION]:
“The use of an LLM (or other AI-tool) for ‘AI assisted copy editing’ purposes does not need to be declared. In this context, we define ‘AI assisted copy editing’ as AI-assisted improvements to human-generated texts for readability and style, and to ensure that the texts are free of errors in grammar, spelling, punctuation and tone. … It does not include generative editorial work and autonomous content creation.” [19]
AI-generated images and videos: “Springer Nature journals are unable to permit its use for publication” given unresolved legal status. [19]
Books: separate book-publishing policies apply; same authorship prohibition.
10. COPE (Committee on Publication Ethics)#
- Primary position statement URL:
https://publicationethics.org/cope-position-statements/ai-author
- Alternative URL:
https://publicationethics.org/guidance/cope-position/authorship-and-ai-tools
(i) Prohibition. COPE issued its position statement on authorship and AI tools on 2023m02d13 (further updates to the page through 2025m02d24). The operative text [QUOTE NEEDS VERIFICATION]:
“AI tools cannot meet the requirements for authorship as they cannot take responsibility for the submitted work. As non-legal entities, they cannot assert the presence or absence of conflicts of interest nor manage copyright and license agreements.” [20]
(ii) Disclosure required. [QUOTE NEEDS VERIFICATION]:
“Authors who use AI tools in the writing of a manuscript, production of images or graphical elements of the paper, or in the collection and analysis of data, must be transparent in disclosing in the Materials and Methods (or similar section) of the paper how the AI tool was used and which tool was used. Authors are fully responsible for the content of their manuscript, even those parts produced by an AI tool, and are thus liable for any breach of publication ethics.” [20]
(iii) Movement / revision.
2023m02d13: COPE position statement issued. [20]
2023–2025: Position statement endorsed by major publishers (BMJ, Wolters Kluwer, SAGE, and others).
2025m02d24: COPE page reviewed/updated [DATE NEEDS VERIFICATION against page footer].
(iv) Carve-outs / grey areas.
COPE is a governance body — it sets principles rather than publisher-specific carve-outs.
Statement does not distinguish writing assistance from research-method use: in both cases authors carry full responsibility.
Some philosophers and bioethicists have publicly contested the “responsibility is required for authorship” premise — see “Daily Nous” 2023m03 commentary and the 2025 PMC article “Responsibility is not required for authorship.” [21] [22]
11. ICMJE (International Committee of Medical Journal Editors)#
- Primary policy URL (AI use by authors):
https://www.icmje.org/recommendations/browse/artificial-intelligence/ai-use-by-authors.html
- Authorship and contributors:
- Full PDF:
(i) Prohibition. ICMJE explicitly states [QUOTE NEEDS VERIFICATION]:
“Chatbots (such as ChatGPT) should not be listed as authors because they cannot be responsible for the accuracy, integrity, and originality of the work, and these responsibilities are required for authorship.” [23]
“Although a chatbot cannot be an author, authors should acknowledge the use of artificial intelligence and chatbots in writing their submission, image creation, or graphical-element production…” [23]
(ii) Disclosure required. [QUOTE NEEDS VERIFICATION]:
“At submission, the journal should require authors to disclose whether they used artificial intelligence (AI)-assisted technologies (such as Large Language Models [LLMs], chatbots, or image creators) in the production of submitted work. Authors who use such technology should describe, in both the cover letter and the submitted work in the appropriate section if applicable, how they used it. For example, if AI was used for writing assistance, describe this in the acknowledgment section. If AI was used for data collection, analysis, or figure generation, authors should describe this use in the methods.” [23]
Two-tier rule (distinctive to ICMJE):
Writing assistance → Acknowledgements.
Data collection / analysis / figure generation → Methods.
(iii) Movement / revision.
2023m05: ICMJE updates Recommendations to address AI for the first time. [24]
2024m01: Further update. [25]
2025: ICMJE 2025 updates respond to rising AI-assisted writing, data transparency, predatory journals, and authorship/peer-review responsibilities. [26]
(iv) Carve-outs / grey areas.
Authors must verify all AI-generated content (facts, references, interpretations). [23]
AI-generated images and figures: covered by Methods disclosure requirement.
AI-assisted translation: must be disclosed.
No carve-out from accountability — humans remain liable.
Cross-cutting movement, dissent, and notable cases#
Early co-author listings (2022m12–2023m01). Several preprints and papers listed ChatGPT as a co-author shortly after its public release in 2022m11. These cases prompted the rapid policy responses from Nature, Science, JAMA, COPE, WAME, and ICMJE in early 2023.
First high-profile retraction for undisclosed ChatGPT use: a paper published 2023m08d09 in Physica Scripta (IOP Publishing) was retracted after a reader spotted the phrase “Regenerate response” — the label of a ChatGPT button — on page 3. The authors acknowledged ChatGPT use; the retraction cited non-disclosure rather than AI use per se. [27]
WAME (World Association of Medical Editors) issued recommendations 2023m01d20 (revised 2023m05d31) aligning with the ICMJE/COPE consensus: chatbots cannot be authors; transparency is required; editors need detection tools. [29]
Internal divergence at NEJM Group: the launch of NEJM AI (2024) with an explicitly LLM-friendly editorial policy creates the first major intra-publisher policy split. [7]
Dissent on mandatory disclosure for writing assistance (2025): Hosseini, Gordijn, Kaebnick, Holmes published “Disclosing generative AI use for writing assistance should be voluntary,” arguing that mandatory disclosure for writing assistance (as opposed to research-method use) can disadvantage non-native English authors and bias editorial decisions. [30]
Dissent on the “responsibility is required for authorship” premise: a 2025 PMC article and 2023 philosophy commentary question whether responsibility is in fact a necessary condition of authorship — challenging the COPE rationale at the philosophical level. [22] [21]
Compliance audit at BMJ Group (2024–2025): 49-journal study of authors’ self-disclosed AI use, presented at the Peer Review Congress — first systematic publisher-internal compliance audit. [12]
Cabanac AI-phrase detection campaign: computer scientist Guillaume Cabanac runs a programme that identifies telltale AI phrases (e.g., “as an AI language model,” “Certainly, here is”) in published literature; Retraction Watch maintains a running tracker. [28]
Consensus position#
Across all eleven publishers and governance bodies surveyed, the convergent position as of 2026m05d12 is unambiguous on three points and unresolved on a fourth:
Convergent (consensus):
AI cannot be listed as an author or co-author on a research paper. Every publisher in this fact-sheet states this explicitly, in similar language, sourced ultimately to the COPE/WAME/ICMJE tri-statement of 2023m01–m05.
AI use must be disclosed. Every publisher requires some form of declaration when AI tools are used in producing the manuscript. The specific location varies (Methods, Acknowledgements, cover letter, dedicated declaration section), but the principle of transparency is universal.
Authors remain fully responsible for all content, including AI-generated portions. No carve-out reduces author liability.
Unresolved / contested:
What level of AI use triggers disclosure is not uniformly defined. Spell-checkers and reference managers are explicitly exempt at Elsevier/JAMA. “AI-assisted copy editing” is exempt at Springer Nature. But the line between copy-editing assistance and generative writing assistance is being actively debated; Hosseini et al. (2025) argue that writing-assistance disclosure should be voluntary, while ICMJE and COPE remain on the mandatory-disclosure side.
Principled arguments the consensus rests on#
The arguments by which publishers justify the prohibition of AI authorship can be enumerated by their explicit sources:
Accountability for the work. Authorship implies responsibility — for accuracy, integrity, and originality. AI tools cannot be held accountable; they cannot respond to critiques, correct the record, or sign-off on a manuscript. Sources: Nature Portfolio / Springer Nature (“attribution of authorship carries with it accountability for the work”); COPE (“AI tools cannot meet the requirements for authorship as they cannot take responsibility”); ICMJE (“they cannot be responsible for the accuracy, integrity, and originality of the work”); JAMA; Cell Press; Elsevier; BMJ; PLOS.
Legal personhood / non-legal entity status. AI tools are not legal persons. They cannot enter copyright or licensing agreements, cannot consent to publication, cannot declare conflicts of interest. Sources: COPE (“As non-legal entities, they cannot assert the presence or absence of conflicts of interest nor manage copyright and license agreements”); WAME (“Under most jurisdictions, an author must be a legal person”).
Failure to satisfy ICMJE authorship criteria. In particular: final approval of the version to be published, and accountability for all aspects of the work — both impossible for AI tools. Sources: ICMJE (origin); WAME; NEJM (by reference to ICMJE); JAMA; BMJ.
Originality and intellectual contribution. Authorship requires substantive intellectual contribution; AI output is derivative and statistically generated from training data. Sources: Science / AAAS; NEJM; Elsevier (“AI output cannot be used directly as manuscript text; rather, AI output can only serve as inspiration … any manuscript needs to represent the authors’ authentic and original contribution”).
Research integrity and transparency. Science depends on transparent methods. Undocumented AI use threatens reproducibility, verifiability, and the chain of evidence. Sources: Nature (originating editorial, 2023m01, “ChatGPT threatens the transparency of methods that are foundational to science”); Science; PLOS; all publishers under the disclosure principle.
Copyright and intellectual-property concerns (specific to generative images). Generative-AI image tools train on potentially copyrighted material; provenance cannot be verified. Sources: Nature (image-policy editorial, 2023m06, “research integrity, consent, privacy and intellectual-property protection”); Springer Nature; Elsevier; NEJM; BMJ; JAMA (no AI-generated images).
Risk of fabrication and error. AI tools can produce confidently-stated but factually wrong output — including non-existent citations. Sources: Cell Press (“AI can generate authoritative-sounding output that can be incorrect, incomplete, or biased”); JAMA; PLOS; NEJM.
Author-as-correspondent obligation. An author must be reachable, must respond to post-publication critique, must be able to defend or correct the work. AI tools cannot fulfil this. Sources: WAME; ICMJE; BMJ; COPE.
Publishers whose policy could not be located#
All eleven publishers and governance bodies surveyed produced retrievable, primary-source policy material via search-snippet evidence. No publisher’s policy was unavailable in this round.
However, the WebFetch denial in this session means none of the
quoted text was directly verified against the live policy page in
this same session. The compiler recommends a second-pass
verification visit to each URL before any quoted operative sentence
is reproduced in a published paper.
Specifically the following quotations carry the [QUOTE NEEDS
VERIFICATION] flag and should be re-checked:
Nature Portfolio operative sentence on LLM authorship.
Science / AAAS operative sentence on AI not meeting authorship criteria.
NEJM exact wording (flagship; relies on ICMJE).
JAMA exact wording of “nonhuman artificial intelligence” provision in instructions-for-authors.
BMJ exact wording of “AI technologies will not be accepted as an author(s)…”
Cell Press operative sentence.
PLOS dedicated-section disclosure language.
Elsevier declaration-statement language.
Springer Nature copy-editing carve-out exact wording.
COPE operative sentences on authorship and on author responsibility.
ICMJE operative sentences on chatbots-not-authors and on two-tier disclosure (cover-letter + Methods/Acknowledgements).
Footnotes#
End of fact-sheet.