Read this first
You have almost certainly not done anything wrong.
Of the 100 highest-ranked journals surveyed in 2023, 87% had published guidance on generative AI and just one prohibited using it to help write a manuscript outright. What nearly all of them prohibited was something different: listing AI as an author. Describe permitted use plainly. A disclosure does not make a prohibited use acceptable.
Ganjavi C, et al. Publishers' and journals' instructions to authors on use of generative artificial intelligence in academic and scientific publishing: bibliometric analysis. BMJ 2024;384:e077192. doi:10.1136/bmj-2023-077192 · PMID 38296328
Where the real lines are
The lines are concrete. Do not invent citations. Do not present generated content as your data. Do not upload an unpublished manuscript while reviewing it unless the journal explicitly authorizes the tool and use. For language help, follow the journal’s disclosure rule and carry on.
If you write in English as a second language, one more thing
Published policy and screening practice can differ. Across 17 cardiovascular journals studied, 15 used AI-detection software and one said so publicly. You should know that these detectors are unreliable, and unreliable in a specific direction: a widely cited study found they wrongly flagged 61% of essays written by non-native English speakers, against almost none by native speakers. That is a defect in the detectors, not in your writing. It is also why a clear, dated note of what you actually did is worth more to you than to anyone else.
Laffaye TA, et al. Artificial intelligence for manuscript writing: policies and implementation in cardiovascular journals. Cardiovasc Diagn Ther 2025. doi:10.21037/cdt-2025-381 · PMID 41216259
Liang W, et al. GPT detectors are biased against non-native English writers. Patterns 2023;4(7):100779. doi:10.1016/j.patter.2023.100779 · PMID 37521038
A policy page, an announcement and a later revision may carry different dates. Springer Nature announced its risk framework on 10 September 2026, although it was already present in the 4 August archive held here. Use the current policy for your decision, and keep a dated copy of the wording you followed.
Springer Nature. Announcement on responsible AI use, 10 September 2026. The framework is also present in the 4 August archive held here. Source
The common cases
Find what you did
"Fixing my English" and "rewriting my text" are not the same thing
Start with what changed. Correcting grammar, tightening phrasing and fixing article use is language editing. Adding arguments, interpretations or conclusions is substantive writing. Translating a draft is a separate activity. Policies draw these boundaries differently; describe the work plainly and use your journal’s rule.
If the distinction is unclear, describe what the tool actually did and check the journal’s rule. A declaration helps readers understand the work, but does not override a prohibition.
Ans M, et al. The Presence and Nature of AI-Use Disclosure Statements in Medical Education Journals. Perspect Med Educ 2026;15:212-225. doi:10.5334/pme.2431 · PMID 41800434
Ordered by how often life scientists actually do these things. Verdicts and placement come from the records further down; the advice and the example wording are ours.
Square brackets are yours to fill. Some are plain facts — the tool, its version, the date. Some are your own words, such as which sections were affected. Where you see a slash inside brackets, pick one. If a tool has no version number you can cite, write the name and the date you used it; that is what the date is for.
I used AI to fix my English, or to tidy up my writingusually nothing to do
Check whether you changed only the language or also the intellectual content. PLOS expressly exempts spelling, grammar and rephrasing for clarity or conciseness from disclosure. If the model introduced arguments or interpretations, use the drafting guidance below.
PLOS. Ethical Publishing Practice — AI tools and technologies. Checked 15 October 2026. Source
[Tool and version] was used for language editing only; no content was generated. The authors are responsible for the final text.
Use this optional wording only when it accurately describes the work and your journal does not require a fuller statement.
If your journal exempts this use but you still want to record it, the acknowledgements are a reasonable home.
Write these in your own words rather than pasting them. A sentence that sounds like you is better than a borrowed one — and more useful to a reader trying to understand what you actually did.
8 of 22 organisations recorded here so far · see the grid
I used AI to draft or substantially rewrite sectionsdeclare it
Check that your journal permits the kind of drafting you used. Then name the tool, version and sections, and explain your own verification. A declaration describes the work; it does not make a prohibited use acceptable.
[Tool and version] was used to draft [which sections], accessed [date]. All generated text was verified against the primary sources and substantially revised by the authors, who take full responsibility for the content.
Name the sections. A statement saying only "AI was used in preparing this manuscript" is the kind editors describe as uninformative — one study found disclosures averaging 43 words and describing the use only superficially.
Ans M, et al. The Presence and Nature of AI-Use Disclosure Statements in Medical Education Journals. Perspect Med Educ 2026;15:212-225. doi:10.5334/pme.2431 · PMID 41800434
- acknowledgementsACS, Frontiers, JAMA Network, PLOS, Royal Society of Chemistry, Sage, Science / AAAS, Wiley
- methodsPLOS, Royal Society of Chemistry, Sage, Science / AAAS, eLife
- dedicated sectionElsevier, JMIR Publications
- cover letterRoyal Society of Chemistry, Science / AAAS
- submission fieldJMIR Publications
Write these in your own words rather than pasting them. A sentence that sounds like you is better than a borrowed one — and more useful to a reader trying to understand what you actually did.
14 of 22 organisations recorded here so far · see the grid
I used AI to find papers or build my reference listdeclare it
Check every reference against the original source before you submit: that it exists, says what you claim, and supports the sentence you cite it for. Declare the assistance where your journal asks for it.
[Tool and version] was used to identify candidate references, accessed [date]. Every reference was subsequently verified against the primary record by the authors.
The second sentence is the one that matters. Say that you checked them.
Check the journal’s disclosure instructions for the required location.
Write these in your own words rather than pasting them. A sentence that sounds like you is better than a borrowed one — and more useful to a reader trying to understand what you actually did.
1 organisation recorded here so far · see the grid
I wrote in my own language and used AI to translatedeclare it
Check the translation and your target journal’s rule. PLOS recommends disclosure but does not require it; other policies ask for a declaration or specify how the translation should be checked.
PLOS. Ethical Publishing Practice — AI tools and technologies. Checked 15 October 2026. Source
The manuscript was drafted in [language] and translated into English using [tool and version] on [date]. The translation was checked by [who checked it, and how], and revised by the authors.
If you have no bilingual colleague to ask — and many people do not — say what you did instead. For example, that you translated the English back into your own language and compared it against your original. Describing a real check honestly is worth more than naming a person who did not do one.
- acknowledgementsJAMA Network, PLOS, Wiley
- dedicated sectionElsevier, Taylor & Francis
- methodsPLOS
Write these in your own words rather than pasting them. A sentence that sounds like you is better than a borrowed one — and more useful to a reader trying to understand what you actually did.
6 of 22 organisations recorded here so far · see the grid
I used AI to write, debug or tidy analysis codedeclare it
Report AI assistance with analysis scripts or software code in your methods, including what the tool did and how you checked the result. PLOS explicitly requires disclosure for generating or developing analysis code. This is research-process use; STM’s separate code-presentation category covers formatting code for publication.
PLOS. Ethical Publishing Practice — AI tools and technologies. Checked 15 October 2026. Source
[Tool and version] was used to assist with writing and debugging analysis code, accessed [date]. All code was reviewed and tested by the authors; analysis scripts are available at [repository].
Report it as software, beside your other tool versions. That is the register it belongs in.
- methodsAIP Publishing, EMBO Press, Elsevier, IOP Publishing, JAMA Network, PLOS, PNAS, Science / AAAS, Wiley
- acknowledgementsIOP Publishing, PLOS, PNAS, Science / AAAS
- cover letterScience / AAAS
Write these in your own words rather than pasting them. A sentence that sounds like you is better than a borrowed one — and more useful to a reader trying to understand what you actually did.
10 of 22 organisations recorded here so far · see the grid
I let an AI agent run analyses, searches or multi-step work for medeclare it
Keep the log: prompts, model and version, dates, tool calls and human checkpoints. A tool name alone cannot explain a multi-step analysis. Follow the journal’s sharing requirements and protect confidential material.
[Tool and version] was used as an agent to [what it executed: run analyses / search literature / generate and run code], between [dates]. Prompts and session transcripts are provided in [supplementary file]. Human verification was performed at [which checkpoints]; outputs that could not be independently reproduced were excluded. The authors take full responsibility for all reported results.
Keep enough detail to trace the work. Share the records required for review and reproducibility, with sensitive material handled appropriately.
- methodsAIP Publishing, EMBO Press, Elsevier, IOP Publishing, JAMA Network, PLOS, PNAS, Science / AAAS, Wiley
- acknowledgementsIOP Publishing, PLOS, PNAS, Science / AAAS
- cover letterScience / AAAS
Write these in your own words rather than pasting them. A sentence that sounds like you is better than a borrowed one — and more useful to a reader trying to understand what you actually did.
10 of 22 organisations recorded here so far · see the grid
I used AI to make a schematic, diagram or graphical abstractcheck before you do it
Check the image category first. Explanatory diagrams, graphical abstracts, cover art and experimental images have different rules. Where permitted, name the tool and describe its use in the required declaration and caption.
Figure [n] was generated using [tool and version] on [date] for illustrative purposes; it does not depict experimental data.
Put it in the figure caption, not only in a statement at the end. That is where a reader meets the image. [n] is your figure number.
- figure captionCell, Elsevier, PLOS
- dedicated sectionCell, Elsevier, Taylor & Francis
- methodsJAMA Network
- submission fieldScience / AAAS
Write these in your own words rather than pasting them. A sentence that sounds like you is better than a borrowed one — and more useful to a reader trying to understand what you actually did.
8 of 22 organisations recorded here so far · see the grid
I used AI to generate or retouch a figure showing my actual datacheck before you do it
A plot made from your data is different from an altered experimental image. Keep the underlying data and reproducible plotting method. Check the journal’s image rules before using AI to retouch a blot, micrograph or scan.
[Tool and version] was used to produce the plots in Figure [n] from the underlying data, accessed [date]. No image content was added, removed or repositioned; source data are available at [location].
Only use this if it is true. Generating a chart from real data is routine; altering an experimental image is not, and that distinction is where retractions happen.
- figure captionJAMA Network, PLOS
- methodsElsevier
Write these in your own words rather than pasting them. A sentence that sounds like you is better than a borrowed one — and more useful to a reader trying to understand what you actually did.
3 of 22 organisations recorded here so far · see the grid
I used AI while peer-reviewing someone else's manuscriptcheck before you do it
Check the journal’s reviewer policy before using a tool. Science allows revision of your own writing only with tools that do not retain inputs or train on them, and requires disclosure in the review. PLOS allows limited support and requires disclosure available to authors. Permission for writing help does not permit uploading a confidential manuscript.
Science journals. Editorial policies — AI and peer review. Checked 15 October 2026. Source
In preparing this review I used [tool and version] to [check my own wording / organise my comments], on [date]. No part of the manuscript was entered into any AI tool. All assessments and conclusions are my own.
Describe only what you actually did. Check the tool’s retention and training settings as well as the journal’s confidentiality rules. PLOS requires reviewer disclosures in the review form and editor disclosures in the decision letter, available to authors.
PLOS. Ethical Publishing Practice — AI tools and technologies. Checked 15 October 2026. Source
- review reportElsevier, Frontiers, MDPI, PLOS, Science / AAAS, Springer Nature
- decision letterPLOS
Write these in your own words rather than pasting them. A sentence that sounds like you is better than a borrowed one — and more useful to a reader trying to understand what you actually did.
12 of 22 organisations recorded here so far · see the grid
I am wondering whether generated content can stand in for data I do not havecheck before you do it
No. This is the one clear prohibition in the dataset, and it is the thing all the other rules exist to protect against. If a figure, a value or a participant did not come from your research, it cannot appear as though it did.
There is no wording for this. It is the only outright prohibition STM's classification carries, and no declaration makes it acceptable.
4 of 22 organisations recorded here so far · see the grid
Things this page does not answer
- Your specific journal. Publisher rules are a starting point. Cell, for example, adds a specific cover-art restriction to Elsevier’s policy. Check the journal instructions as well.
- Your thesis. University rules apply separately. Check the rules for your programme.
- Your co-authors. Agree what the declaration says before submission — it describes the whole manuscript.
- Your funder and preprint server. Their rules apply to different stages of the work. Their records are kept separate from publisher/journal agreement.
- Cover letters and responses to reviewers. Check the named journal. Permission to edit a response does not authorize uploading a confidential review report.
What to actually do
Keep the record. That is the part you cannot recreate.
Start a log, even a rough one
Model and version, rough dates, the prompts, and the session records themselves. Write the declaration while the work is fresh — a first draft takes ten minutes to write. A transcript from eight months ago cannot be rebuilt, and model versions get retired.
Already required, not hypothetical: ISTIC Guideline, JAMA Network, JMIR Publications, Wiley require prompts, versions or archived session records today.
You already used it and kept no record
Write down what you can still remember — the tool, roughly which version, when, and which parts of the work it touched. Do it now: model versions change and chat histories disappear. Then follow the journal’s disclosure instructions. You do not owe anyone an apology in that sentence.
You already found a fabricated reference in your own draft
Catching a fabricated reference before submission is the system working. Remove or replace it, check every remaining reference against the original source, and carry on. If the error has already been published, contact the journal so the record can be put right.
Save a copy of the policy on the day you follow it
Keep the page, its URL and the date you checked it alongside your research notes. Check it again before submission. A dated copy helps you explain which guidance you followed if the wording changes in between.
Check these before you start
Where a use is prohibited outright, the work cannot be fixed at submission. These are the ones worth checking first, with the organisations here that forbid them:
- AI-generated illustrative figuresJAMA Network
- Presenting AI output as original dataElsevier, JAMA Network, PLOS, Royal Society of Chemistry, STM Association
- Using AI when you peer-reviewACS, JAMA Network
Then check which page applies
Authors, reviewers and editors have different responsibilities. Journal-specific rules can add requirements to a publisher policy. Each record below marks the page and audience to cite.
Worked out from the records — never written by hand
Where the policies agree
Each fraction is publishers and journals assessed over publishers and journals in this build. Advisory bodies, funders and preprint servers stay in the records below, outside this comparison. A claim resting on three records is not the same as one resting on thirty, so the denominator is never hidden. Rows move between groups on their own as records are added.
Broad agreement
9 of 16At least three quarters of assessed organisations take the same stance. Exceptions and differences in disclosure requirements are shown below.
- 14/22 organisations assessedUsing AI to draft manuscript text is permitted under conditions and must be declaredacknowledgements, methods, dedicated section
- 12/22 organisations assessedUsing AI when you peer-review is permitted under conditions and disclosure requirements varyAmong organisations allowing this use: 9 require disclosure; 1 do not address disclosurereview report, decision letterexcept ACS, JAMA Network
- 10/22 organisations assessedUsing AI to run the research itself is permitted under conditions and disclosure requirements varyAmong organisations allowing this use: 9 require disclosure; 1 do not address disclosuremethods, acknowledgements, cover letter
- 8/22 organisations assessedUsing AI to polish your language is permitted under conditions and disclosure requirements varyAmong organisations allowing this use: 7 exempt this use; 1 require disclosureconditions vary
- 8/22 organisations assessedAI-generated illustrative figures is permitted under conditions and must be declaredfigure caption, dedicated section, submission fieldexcept JAMA Network, eLife
- 6/22 organisations assessedUsing AI to translate your manuscript is permitted under conditions and disclosure requirements varyAmong organisations allowing this use: 5 require disclosure; 1 encourage disclosureacknowledgements, dedicated section, methodsconditions vary
- 6/22 organisations assessedThe publisher running AI on your manuscript is permitted under conditions and disclosure requirements varyAmong organisations allowing this use: 4 do not address disclosure; 1 require disclosureexcept Wiley
- 5/22 organisations assessedKeeping prompts and session records is permitted under conditions and disclosure requirements varyAmong organisations allowing this use: 2 encourage disclosure; 3 require disclosuremethods, acknowledgements, submission fieldconditions vary
- 4/22 organisations assessedPresenting AI output as original data is prohibited
Genuinely contested
2 of 16Organisations disagree on stance, not merely on conditions. The dissenters are named because they are the useful part.
- 7/22 organisations assessedBeing screened by AI detectors — no majority; organisations are evenly divided
- 6/22 organisations assessedKnowing why you were flagged, and appealing is not addressedexcept Springer Nature, eLife
Too thin to call
3 of 16Fewer than 4 organisations assessed. A rule resting on two records is not a finding, so no claim is made.
- 3/22 organisations assessedAI-generated charts of your real data is permitted under conditions and must be declaredfigure caption, methods
- 2/22 organisations assessedYour paper licensed for AI training is permitted under conditions
- 1/22 organisations assessedUsing AI to find references is discouraged
Not yet assessed
2 of 16No organisation here carries a record on this. Work to do, not a result.
- 0/22 organisations assessedUsing AI to format the data you report
- 0/22 organisations assessedUsing AI to tidy code you report
Nine activities from STM's 2025 classification, then the ones it leaves out
The evidence grid
A dash means we have not assessed that activity. “Not addressed” means the sources we checked say nothing about it. Neither replaces your journal’s instructions. Start with the rule for what you actually did.
Each cell carries two values: whether the use is permitted, and whether it must be declared. silent means we checked and the organisation publishes nothing. — means something weaker: not yet assessed here. Treating those as the same would let unfinished work read as a finding.
| Activity | ACS | AIP Publishing | arXiv | Bernoulli | BMJ | Cambridge University Press | Cell | China GB/Z 290-2026 | COPE | DFG | Directory of Open Access Journals | eLife | Elsevier | EMBO Press | European Research Area living guidelines | European Research Council | Frontiers | ICMJE | IOP Publishing | ISC AI disclosure initiative | ISTIC Guideline | JAMA Network | JMIR Publications | MDPI | National Science Foundation | PLOS | PNAS | Royal Society of Chemistry | Sage | Science / AAAS | Springer Nature | STM Association | Taylor & Francis | World Association of Medical Editors | Wellcome | Wiley |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 1Refinement, correction, editing or formatting the manuscript to improve clarity of language | — not yet assessed | conditionalexempt | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | conditionalexempt | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | permittedexempt | conditionalrequired | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | permittedexempt | — not yet assessed | conditionalexempt | — not yet assessed | permittedexempt | — not yet assessed | — not yet assessed | conditionalexempt | — not yet assessed | — not yet assessed | permittedexempt | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed |
| 2Writing or drafting manuscript content | conditionalrequired | — not yet assessed | conditionalrequired | — not yet assessed | conditionalrequired | conditionalrequired | — not yet assessed | — not yet assessed | conditionalrequired | — not yet assessed | — not yet assessed | conditionalrequired | conditionalrequired | — not yet assessed | — not yet assessed | — not yet assessed | conditionalrequired | conditionalrequired | — not yet assessed | — not yet assessed | conditionalrequired | conditionalrequired | conditionalrequired | — not yet assessed | — not yet assessed | conditionalrequired | — not yet assessed | conditionalrequired | conditionalrequired | conditionalrequired | conditionalrequired | — not yet assessed | — not yet assessed | — not yet assessed | conditionalrequired | conditionalrequired |
| 4Generation, refinement, correction, editing or formatting of images, diagrams or other figures for illustrative purposes only | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | conditionalrequired | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | silentsilent | conditionalrequired | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | prohibitedrequired | — not yet assessed | — not yet assessed | — not yet assessed | conditionalrequired | — not yet assessed | — not yet assessed | — not yet assessed | conditionalrequired | conditionalrequired | — not yet assessed | conditionalrequired | — not yet assessed | — not yet assessed | — not yet assessed |
| 5Translation of manuscript text for the purpose of publishing | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | conditionalrequired | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | conditionalrequired | — not yet assessed | — not yet assessed | — not yet assessed | permittedencouraged | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | conditionalrequired | — not yet assessed | conditionalrequired | — not yet assessed | — not yet assessed | permittedrequired |
| 6Generation, refinement, correction, editing or formatting of visualisations of research data or results | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | conditionalrequired | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | conditionalrequired | conditionalrequired | — not yet assessed | — not yet assessed | — not yet assessed | conditionalrequired | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed |
| 8Assisting with gathering references | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | conditionalrequired | discouragedsilent | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed |
| 9Presentation of any kind of content generated by AI tools as though it were original research data/results from non-machine sources | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | prohibitedsilent | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | prohibitedsilent | — not yet assessed | — not yet assessed | — not yet assessed | prohibitedsilent | — not yet assessed | prohibitedsilent | — not yet assessed | — not yet assessed | — not yet assessed | prohibitedsilent | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed |
| AI used in the research process itself — analysis, code execution, agentic workflows | — not yet assessed | conditionalrequired | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | conditionalrequired | — not yet assessed | — not yet assessed | — not yet assessed | conditionalrequired | conditionalrequired | conditionalencouraged | — not yet assessed | — not yet assessed | conditionalrequired | conditionalrequired | — not yet assessed | — not yet assessed | conditionalrequired | — not yet assessed | — not yet assessed | — not yet assessed | conditionalrequired | conditionalrequired | — not yet assessed | — not yet assessed | conditionalrequired | conditionalsilent | silentsilent | — not yet assessed | — not yet assessed | — not yet assessed | conditionalrequired |
| Use of AI by peer reviewers | prohibitedsilent | — not yet assessed | — not yet assessed | conditionalrequired | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | conditionalrequired | conditionalencouraged | permittedrequired | conditionalrequired | — not yet assessed | — not yet assessed | conditionalsilent | conditionalrequired | conditionalrequired | — not yet assessed | — not yet assessed | conditionalsilent | prohibitedsilent | — not yet assessed | conditionalrequired | conditionalsilent | conditionalrequired | — not yet assessed | — not yet assessed | — not yet assessed | conditionalrequired | conditionalrequired | — not yet assessed | conditionalsilent | — not yet assessed | — not yet assessed | conditionalrequired |
| The publisher's or editor's own AI use on submitted manuscripts | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | conditionalsilent | conditionalsilent | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | conditionalrequired | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | permittedsilent | conditionalsilent | — not yet assessed | — not yet assessed | — not yet assessed | permittedrequired | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | silentsilent |
| Whether AI-text or integrity screening is run, and whether authors are told | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | conditionalsilent | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | silentsilent | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | conditionalsilent | silentsilent | — not yet assessed | — not yet assessed | — not yet assessed | permittedsilent | conditionalsilent | — not yet assessed | — not yet assessed | — not yet assessed | discouragedsilent | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | silentsilent |
| Threshold for review, notification of the author, and route of appeal | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | silentsilent | — not yet assessed | — not yet assessed | conditionalsilent | silentsilent | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | silentsilent | — not yet assessed | — not yet assessed | — not yet assessed | silentsilent | — not yet assessed | — not yet assessed | — not yet assessed | silentsilent | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | conditionalsilent | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | silentsilent |
| Prompts, model versions, session records retained as evidence | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | conditionalencouraged | — not yet assessed | conditionalencouraged | — not yet assessed | conditionalrequired | conditionalrequired | conditionalrequired | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | conditionalencouraged | — not yet assessed | permittedrequired |
| Licensing of author content for AI training, and author consent | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | conditionalsilent | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | conditionalsilent | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed | — not yet assessed |
A ? marks a cell whose verdict is recorded but whose quote is not yet captured. 0 remain, published rather than hidden.
Sources
Organisations, and what to cite
ACS publisher rule with exemptions checked 0d ago
Task-specific rules verified on 11 November 2026. Unassessed activities remain blank.
Cite this
- https://researcher-resources.acs.org/publish/aipolicy
Verified for this PR. Maintainer adjudication remains pending.
What it says, word for word
Writing or drafting manuscript content
Authors may use generative AI tools to assist in preparing their submissions, provided that the use complies with this policy and is disclosed where required below.ACS · ai-policy · checked against the page
Assistance only: no generation of underlying results, scientific reasoning or substantial commentary/review content. Research-method uses have a separate exception.
Use of AI by peer reviewers
peer reviewers remain responsible for all comments and recommendations provided during peer review and must not use generative AI tools to generate, draft, or revise their peer review reports.ACS · ai-policy · checked against the page
Report generation, drafting and revision are prohibited. Confidential material may enter only ACS-authorized integrated tools.
AIP Publishing publisher rule with exemptions checked 0d ago
Task-specific rules verified on 11 November 2026. Unassessed activities remain blank.
Cite this
- https://publishing.aip.org/resources/researchers/policies-and-ethics/ai-policy/
Verified for this PR. Maintainer adjudication remains pending. Selected visible policy passages, captured 11 November 2026. Excerpt, not a full page or raw HTTP response.
What it says, word for word
Refinement, correction, editing or formatting the manuscript to improve clarity of language
The use of AI tools to prepare manuscripts or improve the readability of original content created by the authors is permitted and does not require disclosure.AIP Publishing · ai-policy · checked against a saved browser extract
AI affecting findings or conclusions follows a different disclosure rule.
AI used in the research process itself — analysis, code execution, agentic workflows
AI use should be disclosed when it has the potential to affect the findings or conclusions of an articleAIP Publishing · ai-policy · checked against a saved browser extract
Identify tool/version/manufacturer, purpose and reason; authors validate outputs.
arXiv preprint server rule with exemptions checked 0d ago
Significant tool use should be reported; authors remain responsible. Preprint moderation, not journal peer review.
Cite this
- https://info.arxiv.org/help/moderation/index.html
Verified for this PR. Maintainer adjudication remains pending.
What it says, word for word
Writing or drafting manuscript content
we now include in particular text-to-text generative AI among those that should be reported consistent with subject standards for methodology.arXiv · ai-guidance · checked against the page
Significant tool use should be reported; authors remain responsible. Preprint moderation, not journal peer review.
Bernoulli journal rule with exemptions checked 0d ago
Mathematics/statistics journal. Secure tools must prohibit training on manuscripts. Disclose tools/purposes confidentially to editors; authors agree to safeguarded review use on submission.
Cite this
- https://bernoullisociety.org/publications/bernoulli-journal/policy-on-ai-3
Verified for this PR. Maintainer adjudication remains pending.
What it says, word for word
Use of AI by peer reviewers
Bernoulli allows reviewers and editors to use AI tools to assist in checking technical claims, subject to the confidentiality and data-protection safeguards specified below.Bernoulli · ai-guidance · checked against the page
Mathematics/statistics journal. Secure tools must prohibit training on manuscripts. Disclose tools/purposes confidentially to editors; authors agree to safeguarded review use on submission.
BMJ publisher rule with exemptions checked 0d ago
Task-specific rules verified on 11 November 2026. Unassessed activities remain blank.
Cite this
- https://www.bmj.com/content/ai-use
Verified for this PR. Maintainer adjudication remains pending.
What it says, word for word
Writing or drafting manuscript content
BMJ will consider content where AI technologies are used.BMJ · ai-policy · checked against the page
Declare what, why and how in the contributor section; research uses also belong in Methods.
Whether AI-text or integrity screening is run, and whether authors are told
We may, on a systematic or case by case basis, use screening tools pre-publication to assist with the identification of content generated (in whole or in part) through AI technologies.BMJ · ai-policy · checked against the page
Published screening practice; this does not establish detector accuracy.
Cambridge University Press publisher rule with exemptions checked 0d ago
Task-specific rules verified on 11 November 2026. Unassessed activities remain blank.
Individual journals genuinely differ from this policy — check your specific journal as well.
Cite this
- https://www.cambridge.org/core/services/publishing-ethics/authorship-and-contributorship-journals
Verified for this PR. Maintainer adjudication remains pending.
What it says, word for word
Writing or drafting manuscript content
AI use must be declared and clearly explained in publications such as research papers, just as we expect scholars to do with other software, tools and methodologies.Cambridge University Press · ai-policy · checked against the page
Authors remain accountable; consult the individual journal’s instructions.
Cell journal rule with exemptions checked 0d ago
Task-specific rules verified on 11 November 2026. Unassessed activities remain blank.
Cite this
- https://www.cell.com/cell/information-for-authors/journal-policies
Verified for this PR. Maintainer adjudication remains pending. Selected visible policy passages, captured 11 November 2026. Excerpt, not a full page or raw HTTP response.
What it says, word for word
Generation, refinement, correction, editing or formatting of images, diagrams or other figures for illustrative purposes only
Allows use of generative AI tools in the creation of explanatory images, in line with Elsevier policy.Cell · ai-policy · checked against a saved browser extract
Explanatory figures need tool/version/use. Cover art is prohibited except when AI is integral to the research; general-purpose image tools cannot produce graphical abstracts.
Refinement, correction, editing or formatting the manuscript to improve clarity of language
The declaration does not apply to the use of basic tools, such as tools used to check grammar, spelling, and references.Cell · ai-policy · checked against a saved browser extract
This is an exemption for basic checking, not all drafting.
China GB/Z 290-2026 national body checked 0d ago
Issued 30 October 2026. Official register metadata for the separate national guidance technical document; not ISTIC 3.0. Full normative text remains to be assessed.
Cite this
- https://std.samr.gov.cn/gb/search/gbDetailed?id=5F6B9C9E6B66C75BE06397BE0A0A6C0B
Verified for this PR. Maintainer adjudication remains pending.
COPE normative body prose checked 27d ago
The body most publishers cite, unrevised since 13 February 2023. Its separate discussion document on AI in editorial decisions dates from September 2021, before ChatGPT was released. The position statement places a duty on authors and assigns them liability; it contains no instruction to editors on what to do with a disclosure once received. Live publicationethics.org returns 403 to automated clients, so this record is verified against a dated Internet Archive capture.
Cite this
- Current policy · Archived evidence (live page checked 2026-10-15)
Live publicationethics.org returns 403 to every automated client (Cloudflare). This is a dated Internet Archive capture of the same page. The corresponding live policy was re-read on 15 October 2026; fixed archive captures are retained as evidence and do not establish current live freshness.
Dating and tagging oddities
What it says, word for word
Writing or drafting manuscript content
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 usedCOPE · position-authorship-ai · checked against the page
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.COPE · position-authorship-ai · checked against the page
Methods placement, naming the tool — three years before ICMJE said the same. Note the scope: it already covers collection and analysis of data.
AI used in the research process itself — analysis, code execution, agentic workflows
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 usedCOPE · position-authorship-ai · checked against the page
The statement explicitly extends to "the collection and analysis of data".
DFG funder rule with exemptions checked 0d ago
Funding-proposal reviews. Rules effective 16 April 2026; reviewers disclose use and retain responsibility.
Cite this
- https://www.dfg.de/en/news/news-topics/announcements-proposals/2026/ifr-26-18
Verified for this PR. Maintainer adjudication remains pending.
What it says, word for word
Use of AI by peer reviewers
The guideline allows the use of AI systems in the review process exclusively in a supporting role, linking such use to four central principles: confidentiality, transparency, quality assurance and responsibility.DFG · ai-guidance · checked against the page
Funding-proposal reviews. Rules effective 16 April 2026; reviewers disclose use and retain responsibility.
Directory of Open Access Journals normative body rule with exemptions checked 0d ago
Guidance for journal AI policies; supports disclosed editing, not generated assessments.
Cite this
- https://doaj.org/apply/guide/
Verified for this PR. Maintainer adjudication remains pending.
What it says, word for word
Use of AI by peer reviewers
Peer reviewers and editors should not use generative AI to create their assessments, due to risks such as breaches of confidentiality, superficial and non-specific feedback, bias, hidden prompts , and false information such as fake references; editing and rewriting may be acceptable if disclosed.Directory of Open Access Journals · ai-guidance · checked against the page
Guidance for journal AI policies; supports disclosed editing, not generated assessments.
eLife publisher prose checked 27d ago
Preprint-based review changes what is already public, but does not remove accountability. Author instructions and reviewer guidance are recorded separately.
Cite this
- https://reviewer.elifesciences.org/reviewer-guide/review-process
Since at least 2026-09-11 the URL redirects to the eLife submission system (elife-rp.msubmit.net, reviewer instructions); text unchanged. The public reviewer-guide URL is kept as the citable address.
- https://elife-rp.msubmit.net/html/elife-rp_author_instructions.html
Verified for this PR. Maintainer adjudication remains pending.
What it says, word for word
Use of AI by peer reviewers
While we do not prevent the use of generative artificial intelligence tools within the review process, editors and reviewers remain accountable for their contributions and the contents of their reviews. If generative AI tools are used by editors or reviewers, this must be disclosed.eLife · review-process-ai · checked against the page
Permitted because the confidentiality rationale does not apply to an already-public reviewed preprint — but disclosure is mandatory. An earlier version of this record quoted only the subordinate clause and wrongly recorded disclosure as not addressed.
The publisher's or editor's own AI use on submitted manuscripts
If generative AI is used by editors to source suggestions for potential reviewers, we expect the usual due diligenceeLife · review-process-ai · checked against the page
The guidance addresses AI assistance in sourcing reviewers.
Writing or drafting manuscript content
Whilst such AI tools might be used to aid researchers and authors in doing their work, they cannot be listed as or serve as co-authors of a paper, and their use must be clearly described in the Materials and Methods section.eLife · author-instructions · checked against the page
Authors retain responsibility for the work and for reporting AI outputs.
Threshold for review, notification of the author, and route of appeal
If we receive an allegation of potential research or publication misconduct, we will alert those affected and ask for their response.eLife · author-instructions · checked against the page
General misconduct procedure provides notification and response. The author guide also offers appeals; neither is an AI-detector threshold.
Elsevier publisher rule with exemptions checked 27d ago
Rewrote its journals policy in June 2026, permitting reviewers supportive use of private AI tools. Its BOOKS policy previously used a URL ending 'in-the-review-process' and forbids editors using AI to evaluate material — secondary sources routinely quote that page as if it were the journals policy. Cell Press, an Elsevier imprint, is reported to serve superseded 2023 wording; Cell Press is not yet a record here. Re-read on 2026-09-11 against the text signed off on 2026-08-21: the record had captured five axes of a policy that rules on ten. Added writing, research-process, explanatory-image, data-visualisation and passing-off axes with quotes, and the 18 August 2026 Elsevier Connect announcement as a source.
Individual journals genuinely differ from this policy — check your specific journal as well.
Cite this
Also held, but not the one to cite
- https://www.elsevier.com/about/policies-and-standards/publishing-ethics-books/generative-ai-policies
The slug says 'review process' but the page is Elsevier's BOOKS policy (13 mentions of book/chapter, zero of 'journal' in lower case). Secondary sources routinely quote it as if it were the journals policy. It is not. Recorded until 2026-09-11 as https://www.elsevier.com/about/policies-and-standards/the-use-of-generative-ai-and-ai-assisted-technologies-in-the-review-process, which now redirects here.
- https://www.elsevier.com/connect/updated-generative-ai-policies-for-journals-supporting-responsible-use-while-protecting-trust
Elsevier Connect post announcing the updated journals policy, dated 18 August 2026; the policy page itself states 'Policy updated June 2026'.
Dating and tagging oddities
What it says, word for word
Refinement, correction, editing or formatting the manuscript to improve clarity of language
Basic checks of grammar, spelling and punctuation do not need a declaration statementElsevier · journals-policy · checked against the page
Exemption is withdrawn where changes go beyond basic correction. Specialist disability-related assistive technology used solely for accessibility is also exempt.
Translation of manuscript text for the purpose of publishing
by a person who has a good understanding of both languagesElsevier · journals-policy · checked against the page
Output must be checked by a competent bilingual human.
Use of AI by peer reviewers
Reviewers should not upload a submitted manuscript or any part of it into an AI toolElsevier · journals-policy · checked against the page
AI tools may therefore only be used in a supportive capacity, for example to improve the language and structure of a review report, or to assist with a background literature searchElsevier · journals-policy · checked against the page
Reversed in June 2026. Permitted only with a "private" AI tool as Elsevier defines it; disclosure to authors is mandatory. Uploading the manuscript remains prohibited.
The publisher's or editor's own AI use on submitted manuscripts
Through Elsevier’s editorial systems, editors have access to private AI tools that assist them in performing certain tasks, such as identifying relevant reviewers, matching the scope of submissions to their journal, and detecting duplicate submissions.Elsevier · journals-policy · checked against the page
Authorized private tools within Elsevier editorial systems are distinguished from external supportive tools. Journal and book policies concern different products and activities. The earlier blanket journal-notification sentence is absent from the November snapshot.
Threshold for review, notification of the author, and route of appeal
Authors may be asked to provide documentation of their AI use and the original, unprocessed images for editorial assessment.what is said instead · Elsevier · journals-policy · checked against the page
No threshold for triggering review, no duty to notify the author, and no route of appeal is stated. The quote records what is granted instead: editorial discretion.
Writing or drafting manuscript content
Authors should disclose the use of AI tools for manuscript preparation in a separate AI declaration statement included in their manuscript upon submission.Elsevier · journals-policy · checked against the page
Although AI tools may support your writing process, they must not replace your own scientific judgement, analysis, interpretation or intellectual contribution.Elsevier · journals-policy · checked against the page
Separate AI declaration statement at submission, published with the article; basic grammar and spelling checks exempt, substantive changes to structure or content must be declared. Output must not replace the authors' own judgement, analysis or interpretation.
AI used in the research process itself — analysis, code execution, agentic workflows
Where the use of AI tools forms part of the research design or research methods, the use of the tools must be described in a reproducible manner in the Methods section.Elsevier · journals-policy · checked against the page
Research-method use must be reproducibly described in Methods. The author guidance also names agents and deep research tools for manuscript preparation.
Generation, refinement, correction, editing or formatting of images, diagrams or other figures for illustrative purposes only
Authors may use AI tools to support the creation of certain types of explanatory images for publication, such as flow charts, decision trees, timelines, schematic conceptual illustrations, or experimental workflow diagrams.Elsevier · journals-policy · checked against the page
The use of AI tools should be disclosed in the caption of each image (including the specific tool, version, and how the tool was used) and in the general AI disclosure statement in the article.Elsevier · journals-policy · checked against the page
Explanatory images only (flow charts, decision trees, timelines, schematic illustrations, workflow diagrams); tool, version and use disclosed in each caption and in the AI declaration.
Generation, refinement, correction, editing or formatting of visualisations of research data or results
AI tools may be used to support the creation of data visualizations only when the visual output is directly derived from underlying data.Elsevier · journals-policy · checked against the page
Where AI tools are used in the generation of data visualizations, authors should disclose the name of the model or tool, the version used, and the developer or manufacturer, in the Methods section.Elsevier · journals-policy · checked against the page
Only where the visual is directly derived from underlying data via reproducible methods reported in Methods; tool, version and developer disclosed there.
Presentation of any kind of content generated by AI tools as though it were original research data/results from non-machine sources
AI tools must not be used to fabricate results, invent or alter underlying data, or generate figures that are not faithfully derived from the underlying data and methods used in the research.Elsevier · journals-policy · checked against the page
AI tools must not be used to create or alter images that represent primary observed or experimental data that were not directly obtained in the research.Elsevier · journals-policy · checked against the page
Covers fabricated results, invented or altered data, figures not faithfully derived from data, and primary research images (microscopy, blots, scans, patient images).
EMBO Press publisher rule with exemptions checked 0d ago
Task-specific rules verified on 11 November 2026. Unassessed activities remain blank.
Cite this
- https://link.springer.com/partners/embo-press/editorial-policies
Verified for this PR. Maintainer adjudication remains pending. Selected visible policy passages, captured 11 November 2026. Excerpt, not a full page or raw HTTP response.
What it says, word for word
Refinement, correction, editing or formatting the manuscript to improve clarity of language
Use of AI programs should be primarily restricted to improve the style, accessibility or quality of human generated text and images, including schemata and data figures.EMBO Press · ai-policy · checked against a saved browser extract
Declare affected parts and tools in the published article.
AI used in the research process itself — analysis, code execution, agentic workflows
For data based on original experimentation or meta-analysis of the literature, AI programs can be used in the collection and analysis of data if this is documented in the Methods section of the paper.EMBO Press · ai-policy · checked against a saved browser extract
Label synthetic data; never misrepresent experimental findings.
European Research Area living guidelines normative body rule with exemptions checked 0d ago
Non-binding fourth edition, October 2026. Substantial use, human responsibility, checkpoints and reproducibility; not a journal submission rule.
Cite this
- https://research-and-innovation.ec.europa.eu/document/download/2b6cf7e5-36ac-41cb-aab5-0d32050143dc_en
Verified for this PR. Maintainer adjudication remains pending.
What it says, word for word
AI used in the research process itself — analysis, code execution, agentic workflows
Researchers disclose which AI tools or systems were used substantially10 in their research or reporting processes and describe what they contributed or were used for.European Research Area living guidelines · ai-guidance · checked against the page
Non-binding fourth edition, October 2026. Substantial use, human responsibility, checkpoints and reproducibility; not a journal submission rule.
European Research Council funder rule with exemptions checked 0d ago
Funding review: no proposal summarization, scientific-merit assessment or draft evaluations; external uploads prohibited.
Cite this
- https://erc.europa.eu/news-events/news/erc-clarifies-limits-ai-use-grant-evaluation
Verified for this PR. Maintainer adjudication remains pending.
What it says, word for word
Use of AI by peer reviewers
Reviewers may use AI tools to improve the language of their reports or to search for general information, provided no proposal content or personal data is shared and no assessment is delegated.European Research Council · ai-guidance · checked against the page
Funding review: no proposal summarization, scientific-merit assessment or draft evaluations; external uploads prohibited.
Frontiers publisher rule with exemptions checked 0d ago
Task-specific rules verified on 11 November 2026. Unassessed activities remain blank.
Cite this
- https://www.frontiersin.org/guidelines/policies-and-publication-ethics
Verified for this PR. Maintainer adjudication remains pending.
What it says, word for word
Writing or drafting manuscript content
If AI tools have been used to generate main text, then this must be clearly disclosed in the acknowledgments.Frontiers · ai-policy · checked against the page
Check accuracy and originality; authors remain responsible.
Use of AI by peer reviewers
To support peer reviewers, we provide an internal AI tool within the review forum, operating in a secure, closed environment.Frontiers · ai-policy · checked against the page
Internal support does not make decisions. External use needs advance editorial discussion and, if approved, disclosure to authors; no manuscript upload to external tools.
Prompts, model versions, session records retained as evidence
We encourage authors to upload all input prompts provided to a generative AI technology and outputs received from a generative AI technology in the supplementary files for the manuscript.Frontiers · ai-policy · checked against the page
Supplementary prompts and outputs are encouraged, not universally mandatory.
ICMJE normative body prose checked 27d ago
Updated January 2026 with a whole new Section V. Files AI structurally INSIDE contributorship (II.A.4, directly after Non-Author Contributors) and NOT under conflicts of interest — which refutes the naive claim that guidance treats AI use as a competing interest. Section V.C is the
Cite this
- https://www.icmje.org/recommendations/browse/artificial-intelligence/ai-use-by-authors.html
- https://www.icmje.org/recommendations/browse/artificial-intelligence/ai-use-by-editors.html
- https://www.icmje.org/recommendations/browse/artificial-intelligence/ai-use-by-reviewers.html
Verified for this PR. Maintainer adjudication remains pending.
Also held, but not the one to cite
- https://www.icmje.org/recommendations/browse/manuscript-preparation/preparing-for-submission.html
- https://www.icmje.org/recommendations/browse/roles-and-responsibilities/defining-the-role-of-authors-and-contributors.html
Dating and tagging oddities
What it says, word for word
Writing or drafting manuscript content
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.ICMJE · authors-contributors · checked against the page
Disclosure duty at submission, in the cover letter AND in the work itself.
AI used in the research process itself — analysis, code execution, agentic workflows
including the tool used, version, and prompts where applicable.ICMJE · preparing-submission · checked against the page
The tool-and-version instruction already exists here. But the required fields are dimensioned for a citation, not a trace — adequate for a paragraph, meaningless for forty tool calls. The document contains zero occurrences of agent, agentic or autonomous, against four of "chatbot", under a Section V titled "Use of AI in Publishing".
The publisher's or editor's own AI use on submitted manuscripts
Journals should have a policy for the use of AI in the review process and should make all editors, reviewers, and authors aware of this journal policy.ICMJE · ai-editors · checked against the page
Section V.C. The first obligation anywhere on journals to disclose their own AI use — which converts the author/publisher asymmetry from an ethical complaint into a measurable compliance gap.
Threshold for review, notification of the author, and route of appeal
Editors should be vigilant that authors may not disclose AI use when queried.what is said instead · ICMJE · ai-editors · checked against the page
Nondisclosure of AI use may require corrective action and may be construed as misconduct in some circumstanceswhat is said instead · ICMJE · ai-authors · checked against the page
No threshold for triggering review, no duty to notify the author, and no route of appeal is stated. The quote records what is granted instead: editorial discretion.
Use of AI by peer reviewers
Reviewers must adhere to the journal’s stated policy on the use of AI or request permission from the journal prior to using AI technology to facilitate their review.ICMJE · ai-reviewers · checked against the page
Follow the journal policy or request permission; maintain confidentiality, disclose use to the journal and validate the output.
IOP Publishing publisher rule with exemptions checked 0d ago
Task-specific rules verified on 11 November 2026. Unassessed activities remain blank.
Cite this
- https://publishingsupport.iopscience.iop.org/ethical-policy-journals/
Verified for this PR. Maintainer adjudication remains pending.
Also held, but not the one to cite
- https://ioppublishing.org/legal/textanddataminingpolicy/
Verified for this PR. Maintainer adjudication remains pending.
What it says, word for word
AI used in the research process itself — analysis, code execution, agentic workflows
GenAI tools may be used as part of the research process, including to support research ideation, methodology, analysis where their use is methodologically appropriate, transparent and reproducible.IOP Publishing · ai-policy · checked against the page
Material uses require an AI statement in Acknowledgements and research-method detail in Methods.
Prompts, model versions, session records retained as evidence
Authors are encouraged, and may be required in fields where AI is central to the research method, to provide relevant prompts, prompt logs, model settings, outputs and validation materials as supplementary data where this is necessary to support transparency, reproducibility and assessment of the work.IOP Publishing · ai-policy · checked against the page
Conditional requirement where AI is central; retain the distinction from a universal prompt-publication mandate.
Licensing of author content for AI training, and author consent
AI and/or T&DM for commercial purposes is not permitted unless expressly agreed in writing by IOP Publishing.IOP Publishing · tdm-ai · checked against the page
Publisher-stated access and reuse terms. Scope depends on content, license and purpose; this is not a finding about author consent or the reach of legal exceptions.
ISC AI disclosure initiative normative body checked 0d ago
Vancouver Standard initiative with COPE, STM and GYA. Consultation in progress, not an adopted standard. No permission verdicts are inferred from survey questions.
Cite this
- https://council.science/our-work/ai-disclosure-in-research/
Verified for this PR. Maintainer adjudication remains pending.
ISTIC Guideline national body rule with exemptions checked 27d ago
Version 3.0 (December 2025) is publisher-co-signed guidance, not the separate Chinese national document. GB/Z 290-2026 was issued on 30 October 2026; its full normative text still needs separate assessment. Disclosure, reviewer support and research-process rules below come from the English v3.0 operative sections.
Cite this
Also held, but not the one to cite
What it says, word for word
Assisting with gathering references
Literature research: AIGC tools can be used to collect reference literature based on keywords or topics, classi- fy and review the literature, summarize the conclusions, and provide references for researchers;ISTIC Guideline · guideline-v3-en · checked against the page
Section 4.1.1: verify every source and retain human-led selection. Deep research outputs need the same checks.
Use of AI by peer reviewers
During the manuscript review process, reviewers must determine whether AIGC tools can be used based on the policies of the commissioning party (e.g., journals or publishers). If journal policies explicitly prohibit their use, AIGC tools must not be employed.ISTIC Guideline · guideline-v3-en · checked against the page
Section 4.2.4.2: follow the commissioning journal. Support must enhance the reviewer’s own comments; the text also forbids processing content under review and restricts public-tool uploads.
Writing or drafting manuscript content
Also, AIGC can be used to summa- rize other scholarly publications during the research process or to generate the literature review section. However, it should not be used to write the entire text, interpret data, or draw scienti�ic conclusions.ISTIC Guideline · guideline-v3-en · checked against the page
Section 4.1.4 allows literature-review assistance, not whole-paper writing or replacement of scientific judgment. Section 4.2.3 requires disclosure.
Generation, refinement, correction, editing or formatting of visualisations of research data or results
Statistical chart and diagram format processing and optimization: When processing and optimizing statisti- cal charts, researchers can use AIGC tools to help adjust chart styles according to the amount of data and speci�ic needs.ISTIC Guideline · guideline-v3-en · checked against the page
Section 4.1.3 permits chart formatting and data-grounded examples. Experimental images must come from research; graphical abstracts are prohibited and cover art needs prior permission.
Prompts, model versions, session records retained as evidence
The user; the arti�icial intelligence technology or system (version number required); the time and date of use; prompts and questions used for content generation or optimization;ISTIC Guideline · guideline-v3-en · checked against the page
Section 4.2.3 specifies user, tool/version, time/date, prompts, affected parts and ideas. Its template describes attached records; section 4.3.1 addresses supplementary archiving. This is more specific than a tool-name declaration, not a blanket demand to publish every session.
Whether AI-text or integrity screening is run, and whether authors are told
The detection results should be used as an auxiliary supporting basis in a comprehensive evaluationISTIC Guideline · guideline-v3-en · checked against the page
Section 4.3.2: account for false positives and evaluate detections in context. This is guidance on detection, not evidence that ISTIC screens submissions.
JAMA Network publisher per task matrix checked 27d ago
Publisher of JAMA and the JAMA Network specialty journals (American Medical Association). One instructions-for-authors page serves every title, so the AI rules are network-wide. Revised guidance was published as an editorial on 2026-08-10 (Flanagin, Perlis, Bibbins-Domingo; doi:10.1001/jama.2026.16613, Crossref-verified 2026-09-11) with a table grading fourteen use types; the instructions page was updated 2026-07-29 to add the prohibition on drafting Opinion pieces, Letters and Comments. Alongside Science / AAAS it is one of two publishers here ruling on AI-drafted responses to reviewers (not permitted). Live jamanetwork.com returns 403 to automated clients; both sources are dated Internet Archive captures.
Cite this
- Current policy · Archived evidence (live page checked 2026-10-15)
Live jamanetwork.com returns 403 to automated clients. Dated Internet Archive capture; the page states 'Last Updated: July 29, 2026'. The previous capture (2026-07-01) lacks the Opinion/Letters/Comments prohibition, so that rule dates from July 2026. The corresponding live policy was re-read on 15 October 2026; fixed archive captures are retained as evidence and do not establish current live freshness. The live instructions show Last Updated: September 8, 2026; the July date above describes the retained capture. The recorded AI provisions remain present.
Also held, but not the one to cite
- Current policy · Archived evidence (live page checked 2026-10-15)
Flanagin A, Perlis RH, Bibbins-Domingo K. Updated Guidance for Author Use of AI in Medical Publication. JAMA, published online 2026-08-10, doi:10.1001/jama.2026.16613. Carries the acceptable-use table and the Box reproduced in the instructions. Dated Internet Archive capture. The corresponding live policy was re-read on 15 October 2026; fixed archive captures are retained as evidence and do not establish current live freshness.
What it says, word for word
Writing or drafting manuscript content
Use of AI, LLMs, chatbots, or similar technologies to assist with content creation, revision, and formatting must be reported in the Acknowledgment section.JAMA Network · instructions-for-authors · checked against the page
Use of AI, LLM, and chatbots to draft Opinion manuscripts, Letters to the Editor, and Online Comments is not permitted.JAMA Network · instructions-for-authors · checked against the page
Not permitted at all for Opinion manuscripts, Letters to the Editor and online Comments ('human insights are prioritized'). Acknowledgment must name tool, version, manufacturer, dates of use, what it was used on, and confirm author responsibility.
Refinement, correction, editing or formatting the manuscript to improve clarity of language
Note this guidance does not apply to basic tools for checking grammar or spelling.JAMA Network · instructions-for-authors · checked against the page
Use of AI tools and software to check grammar and spelling or for translation is permitted, but only after the initial text is written by humans.JAMA Network · editorial-2026-08 · checked against the page
Only after the initial text is written by humans.
Translation of manuscript text for the purpose of publishing
Translation Yes, provided the use is fully described in the Acknowledgment section and authors check for accuracy and take full responsibility for all contentJAMA Network · editorial-2026-08 · checked against the page
Table row; warns against entering content into public translation tools.
Assisting with gathering references
AI, LLM, and chatbots should not be used to generate or format references as these tools may cause problems with the accuracy and integrity of cited references; standard reference managers may be used instead.JAMA Network · instructions-for-authors · checked against the page
The editorial's table grades this 'Discouraged'; the instructions say 'should not'.
Generation, refinement, correction, editing or formatting of images, diagrams or other figures for illustrative purposes only
Submission of clinical images and clinical illustrations created or manipulated by AI, LLM, and chatbots is not permitted, unless part of formal research design or methods, which must be fully disclosed.JAMA Network · instructions-for-authors · checked against the page
Scope is clinical images and clinical illustrations (and clinical video/audio); the sole exception is use as part of formal research design or methods, which must be fully disclosed. Non-clinical explanatory figures are not addressed.
Generation, refinement, correction, editing or formatting of visualisations of research data or results
Use of AI to generate graphic visualization of research data or results is permitted, provided that the AI or software (name and version number) used is included in the figure legend.JAMA Network · editorial-2026-08 · checked against the page
Name and version number of the AI or software in the figure legend.
Presentation of any kind of content generated by AI tools as though it were original research data/results from non-machine sources
Presenting AI-generated data as if it were original research from non-AI sources NoJAMA Network · editorial-2026-08 · checked against the page
Table row in the editorial.
AI used in the research process itself — analysis, code execution, agentic workflows
Research Yes, provided the use is fully described in the Methods section and reporting guidance is followed and authors take full responsibility for all contentJAMA Network · editorial-2026-08 · checked against the page
Describe how AI was used for specific aspects of the study (eg, to generate or refine study hypotheses, assist in the generation of a list of adjustment variables, create graphs to show visual relationships).JAMA Network · instructions-for-authors · checked against the page
Follow the relevant reporting guideline; describe each use in Methods; for LLM studies give platform, version, manufacturer, dates and prompts; for ML give training/validation data, metrics, bias and subgroup analyses; data-sharing statement including code.
Prompts, model versions, session records retained as evidence
For studies using LLMs, provide the name of the platform or program, tool, version, and manufacturer; specify dates and prompt(s) used and their sequence and any revisions to prompts in response to initial outputs.JAMA Network · instructions-for-authors · checked against the page
For research use: prompts, their sequence and revisions, dates, platform and version in Methods. For manuscript preparation: tool, version, manufacturer and dates in the Acknowledgment. No requirement to archive session records.
Use of AI by peer reviewers
Entering an author’s manuscript or the reviewer’s summary of the manuscript into an external AI tool or LLM is a violation of confidentialityJAMA Network · editorial-2026-08 · checked against the page
Stated in the author-facing table; JAMA Network's separate reviewer instructions are not held here. The same table also bars AI-drafted author responses to reviewer and editor comments.
Threshold for review, notification of the author, and route of appeal
If incomplete disclosure or failure to disclose use of AI is detected during the editorial process or after publication, content may be rejected or subject to postpublication corrective action.what is said instead · JAMA Network · instructions-for-authors · checked against the page
A sanction is stated (rejection or post-publication correction) but no threshold for suspicion, no duty to notify the author and no route of appeal.
JMIR Publications publisher rule with exemptions checked 0d ago
Task-specific rules verified on 11 November 2026. Unassessed activities remain blank.
Cite this
- https://support.jmir.org/hc/en-us/articles/13387268671771-JMIR-Publications-Editorial-Policy-on-the-use-of-generative-AI-during-manuscript-preparation
Verified for this PR. Maintainer adjudication remains pending.
What it says, word for word
Prompts, model versions, session records retained as evidence
We ask authors to keep the complete original prompts, responses, and transcripts on file and to submit all generative AI conversations used in the preparation of the manuscript as a Multimedia Appendix, so reviewers and readers can assess the extent of AI-generated content in the manuscript.JMIR Publications · ai-policy · checked against the page
Declare at submission and each resubmission. Retain the complete prompts, responses and conversations, with supplementary materials as instructed.
Writing or drafting manuscript content
We do allow the use of generative AI in ideation, but ask for full disclosure on the extent of use of generative AI in the ideation or writing process.JMIR Publications · ai-policy · checked against the page
AI-generated text must be identified as such. The October 2026 update replaces acknowledgments with a dedicated disclosure statement.
MDPI publisher rule with exemptions checked 0d ago
Task-specific rules verified on 11 November 2026. Unassessed activities remain blank.
Cite this
- https://www.mdpi.com/ethics
Verified for this PR. Maintainer adjudication remains pending. Selected visible policy passages, captured 11 November 2026. Excerpt, not a full page or raw HTTP response.
What it says, word for word
Refinement, correction, editing or formatting the manuscript to improve clarity of language
The use of GenAI tools for the purposes of text editing (e.g., grammar, structure, spelling, punctuation, and formatting) is not covered by this policy and does not need to be declared.MDPI · ai-policy · checked against a saved browser extract
Editing exemption includes structure.
Use of AI by peer reviewers
Some limited use to improve the written quality of a peer review report, such as checking grammar, structure, spelling, punctuation, and formatting, may be acceptable but should be disclosed upon submission of the peer review report.MDPI · ai-policy · checked against a saved browser extract
Limited editing only; no manuscript or review communications may be uploaded.
National Science Foundation funder rule with exemptions checked 0d ago
Proposal and review material cannot be uploaded to non-approved tools. Public-information support is distinct from delegation of evaluation.
Cite this
- https://www.nsf.gov/policies/ai/merit-review
Verified for this PR. Maintainer adjudication remains pending.
What it says, word for word
Use of AI by peer reviewers
NSF reviewers may share publicly available information with current generation generative AI tools.National Science Foundation · ai-guidance · checked against the page
Proposal and review material cannot be uploaded to non-approved tools. Public-information support is distinct from delegation of evaluation.
PLOS publisher rule with exemptions checked 27d ago
The AI section changed between the successful fetches of 11 September and 1 October 2026; the 15 October fetch matches the 1 October wording. It now expressly exempts language improvements from disclosure, recommends disclosure for translation, and addresses analysis code, figures and AI detection. Reviewer and editor disclosures must be available to authors. The policy distinguishes disclosure from undisclosed or unacceptable use; disclosure itself is not grounds for an adverse decision.
Cite this
- https://journals.plos.org/plosone/s/ethical-publishing-practice
No revision date is stated. A text change is bounded by the 11 September and 1 October 2026 fetches; it is not dated precisely. Re-read on 15 October.
What it says, word for word
Refinement, correction, editing or formatting the manuscript to improve clarity of language
We do not require disclosure of AI use for language improvements, such as to correct spelling or grammar, or to rephrase content for clarity or conciseness.PLOS · ethical-publishing-practice · checked against the page
Exemption covers spelling, grammar and rephrasing for clarity or conciseness; it does not cover generating new intellectual content.
Writing or drafting manuscript content
AI tools may be used to support development or revision of a PLOS submission.PLOS · ethical-publishing-practice · checked against the page
Any AI use to generate or revise content for a PLOS submission, or to support research reported in a PLOS article, must be clearly disclosed as outlined in the AI use disclosures section of this policy.PLOS · ethical-publishing-practice · checked against the page
For AI use by authors, disclosures should be in the body of the article, either in a dedicated section of the Methods, in the Acknowledgement section, or in the relevant figure legend(s).PLOS · ethical-publishing-practice · checked against the page
Authors retain responsibility for accuracy and their own ideas. Name the tool, explain its use and affected content, and describe how outputs were checked. Methods and Acknowledgements are alternative locations; figure-specific use belongs in the legend.
Translation of manuscript text for the purpose of publishing
If an AI tool has been used for language translation, we recommend this be disclosed in order to preempt concerns, but this is not strictly required.PLOS · ethical-publishing-practice · checked against the page
Disclosure is recommended to preempt concerns but is expressly not compulsory.
Generation, refinement, correction, editing or formatting of images, diagrams or other figures for illustrative purposes only
For example, authors can use AI-generated figures for illustrative purposes or to summarize results, but should be aware that depending on the AI tool’s terms and conditions any data uploaded to the tool may be ingested into the AI tool’s training dataset and exposed to future users.PLOS · ethical-publishing-practice · checked against the page
Any AI use for figure preparation should be clearly disclosed in the figure legend.PLOS · ethical-publishing-practice · checked against the page
Illustrative figures must not misrepresent primary research, introduce invalid content or put sensitive data at risk. Check accuracy, copyright and licensing; disclose use in the legend.
Generation, refinement, correction, editing or formatting of visualisations of research data or results
For example, authors can use AI-generated figures for illustrative purposes or to summarize results, but should be aware that depending on the AI tool’s terms and conditions any data uploaded to the tool may be ingested into the AI tool’s training dataset and exposed to future users.PLOS · ethical-publishing-practice · checked against the page
Any AI use for figure preparation should be clearly disclosed in the figure legend.PLOS · ethical-publishing-practice · checked against the page
Figures summarizing results are allowed only when valid and faithful to the research outputs; disclosure belongs in the legend. This does not permit fabricating or misrepresenting primary data.
Presentation of any kind of content generated by AI tools as though it were original research data/results from non-machine sources
The use of AI tools and technologies to fabricate or otherwise misrepresent primary research data is unacceptable and will be considered by PLOS to be a breach of publication ethics.PLOS · ethical-publishing-practice · checked against the page
Fabrication or misrepresentation of primary research data is a breach of publication ethics.
AI used in the research process itself — analysis, code execution, agentic workflows
Any AI use to generate or revise content for a PLOS submission, or to support research reported in a PLOS article, must be clearly disclosed as outlined in the AI use disclosures section of this policy.PLOS · ethical-publishing-practice · checked against the page
Do I need to disclose AI use for generation or development of analysis scripts or software code? Yes.PLOS · ethical-publishing-practice · checked against the page
Includes development of analysis scripts or software code. Name the tools, identify affected study components and explain output validation. Code generation belongs to this research-process axis, which STM excludes from code formatting.
Use of AI by peer reviewers
AI tools can only be used in peer review for purposes that (i) do not place the confidentiality of manuscript content or data at risk, AND (ii) would not misrepresent the reviewer’s or editor’s own intellectual contributions.PLOS · ethical-publishing-practice · checked against the page
For AI use by editors or reviewers, the disclosure should be in the review form or decision letter so that the information about AI use is available to authors when they receive their decision letter.PLOS · ethical-publishing-practice · checked against the page
Only support that preserves confidentiality and the reviewer’s own intellectual contribution is allowed. AI may not summarize or evaluate submissions. Reviewers disclose in the review form; editors in the decision letter, available to authors.
The publisher's or editor's own AI use on submitted manuscripts
PLOS uses AI tools to support submission checks, screening, and integrity investigations, and we may also integrate AI tools to further support our publishing processes, communications, and author services.PLOS · ethical-publishing-practice · checked against the page
PLOS describes vetted AI tools for submission checks, screening and integrity investigations. It does not promise an article-specific account of which internal tools ran. Editor decision-letter disclosure is recorded separately under reviewer use.
Whether AI-text or integrity screening is run, and whether authors are told
We may use available AI detection tools in article-level checks or assessments, and we follow up if editors, reviewers, staff, or readers report concerns about undisclosed AI use.PLOS · ethical-publishing-practice · checked against the page
The policy expressly allows AI detection at article level; it does not say every submission is screened or promise individual notification.
Threshold for review, notification of the author, and route of appeal
Will AI use disclosures negatively affect my article’s review or editorial decision? No.what is said instead · PLOS · ethical-publishing-practice · checked against the page
If you have concerns about your submission’s decision or review comments, please contact the journal office.what is said instead · PLOS · ethical-publishing-practice · checked against the page
The AI section offers a journal-office contact for decision or review concerns, and says disclosure itself will not adversely affect the decision. It does not specify an investigation threshold, a notification duty or a formal appeal procedure. These quotations document the limited guidance that is provided.
PNAS journal rule with exemptions checked 0d ago
Task-specific rules verified on 11 November 2026. Unassessed activities remain blank.
Cite this
- https://www.pnas.org/author-center/editorial-and-journal-policies
Verified for this PR. Maintainer adjudication remains pending. Selected visible policy passages, captured 11 November 2026. Excerpt, not a full page or raw HTTP response.
What it says, word for word
AI used in the research process itself — analysis, code execution, agentic workflows
Use of AI and generative AI software, such as large language models (e.g., ChatGPT), during the research process must be disclosed in the Materials and Methods section (or Acknowledgments, if no Materials and Methods section is available) of the manuscript.PNAS · ai-policy · checked against a saved browser extract
Identify tool/model and verify all outputs.
Whether AI-text or integrity screening is run, and whether authors are told
Pangram to detect content generated by AIPNAS · ai-policy · checked against a saved browser extract
PNAS publicly names Pangram for AI-text detection and iThenticate for similarity checks. Naming a tool is not a claim of accuracy.
The publisher's or editor's own AI use on submitted manuscripts
Prophy to assist matching submissions to editors and potential reviewers, and Web of Science Reviewer Locator to identify potential reviewers.PNAS · ai-policy · checked against a saved browser extract
Publicly described reviewer/editor matching; no inferred per-manuscript disclosure requirement.
Royal Society of Chemistry publisher rule with exemptions checked 0d ago
Task-specific rules verified on 11 November 2026. Unassessed activities remain blank.
Cite this
- https://www.rsc.org/publishing/journals/processes-and-policies/author-responsibilities
Verified for this PR. Maintainer adjudication remains pending. Selected visible policy passages, captured 11 November 2026. Excerpt, not a full page or raw HTTP response.
What it says, word for word
Writing or drafting manuscript content
If GenAI tools have been used when generating any part of the manuscript (for example when drafting text, translating content, formatting data, creating or editing figures, visualising results, refining code, or compiling references) this must be declared at submission within the cover letter.Royal Society of Chemistry · ai-policy · checked against a saved browser extract
Methods describes use and prompts; acknowledgments identifies the tool/model. Copyediting is exempt.
Presentation of any kind of content generated by AI tools as though it were original research data/results from non-machine sources
We do not permit the presentation of any kind of content generated by AI tools as though it were original research data/results from nonmachine sources.Royal Society of Chemistry · ai-policy · checked against a saved browser extract
Generated content cannot stand in for nonmachine research data.
Sage publisher rule with exemptions checked 0d ago
Task-specific rules verified on 11 November 2026. Unassessed activities remain blank.
Cite this
- https://www.sagepub.com/journals/publication-ethics-policies/artificial-intelligence-policy
Verified for this PR. Maintainer adjudication remains pending.
Also held, but not the one to cite
- https://www.sagepub.com/tdm-ai-policy
Verified for this PR. Maintainer adjudication remains pending.
What it says, word for word
Refinement, correction, editing or formatting the manuscript to improve clarity of language
AI tools that make suggestions to improve or enhance your own work, such as tools to improve language, grammar, or structure, are considered assistive AI tools and do not require disclosure by authors or reviewers.Sage · ai-policy · checked against the page
Assistive editing of the author’s own work.
Writing or drafting manuscript content
The primary or partial use of AI tools and/or LLMs that produce content such as references, text, images, or any other content that directly impacts the research methodology, analysis, results and/or conclusions must be disclosed upon submission so the editorial team can evaluate the content generated.Sage · ai-policy · checked against the page
Verify outputs and sources. Disclosing use does not authorize prohibited activities.
Licensing of author content for AI training, and author consent
If you wish to use Sage journals content for any AI uses including generative AI you should request permissionSage · tdm-ai · checked against the page
Publisher-stated access and reuse terms. Scope depends on content, license and purpose; this is not a finding about author consent or the reach of legal exceptions.
Science / AAAS publisher per task matrix checked 27d ago
Reviewer writing support and image exceptions are recorded from the held policy. The linked AI-use matrix needs a separate full capture before its editorial-disclosure rules are assigned. General integrity procedures must be assessed separately from AI-specific thresholds.
Cite this
- Current policy · Archived evidence (live page checked 2026-10-15)
Live science.org returns 403 to every automated client. This is a dated Internet Archive capture. The corresponding live policy was re-read on 15 October 2026; fixed archive captures are retained as evidence and do not establish current live freshness.
What it says, word for word
Use of AI by peer reviewers
Reviewers may not enter any part of the manuscript into an LLM or other AI tool because this could breach the confidentiality of the manuscript.Science / AAAS · editorial-policies · checked against the page
However, reviewers may use AI tools to revise their own writing, provided that the system does not save or use inputs to train the model and that such use is transparently declared in the review.Science / AAAS · editorial-policies · checked against the page
Reviewers may revise their own writing if the tool neither saves inputs nor uses them for training, and they disclose this in the review. Entering any manuscript content remains prohibited. This corrects the former blanket prohibition in this record; the exception was already present in the stored source.
Generation, refinement, correction, editing or formatting of images, diagrams or other figures for illustrative purposes only
AI-generated images and other multimedia are not permitted in the Science journals without explicit permission from the editors.Science / AAAS · editorial-policies · checked against the page
Prior editor permission required; exceptions case-by-case and disclosed at submission.
Writing or drafting manuscript content
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 methods section or acknowledgments section of the manuscriptScience / AAAS · editorial-policies · checked against the page
Location and detail depend on use case. Authors verify accuracy, sources and bias; the separately linked matrix needs its own capture.
AI used in the research process itself — analysis, code execution, agentic workflows
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 methods section or acknowledgments section of the manuscriptScience / AAAS · editorial-policies · checked against the page
Location and detail depend on use case. Authors verify accuracy, sources and bias; the separately linked matrix needs its own capture.
Springer Nature publisher risk tiered checked 27d ago
Role-specific writing, research and editor guidance supplements the Nature umbrella and reviewer sources. The author declaration wording is more explicit than the umbrella green-tier wording; use the role-specific instructions.
Cite this
- Current policy · Archived evidence (live page checked 2026-10-15)
Live nature.com returns a Cloudflare 'Client Challenge' interstitial with HTTP 200 to automated clients — a 3KB block page, not the policy. This is a dated Internet Archive capture of the real page. The corresponding live policy was re-read on 15 October 2026; fixed archive captures are retained as evidence and do not establish current live freshness.
- https://www.springernature.com/de/policies/editorial-policies/using-ai-in-peer-review
Direct HTTP returns a challenge page. The stored text is a browser DOM excerpt of the visible risk-framework table, captured on 15 October 2026; it is not raw HTTP bytes or a full-page snapshot.
- https://www.springernature.com/gp/policies/editorial-policies/ai-manuscript-preparation
Verified for this PR. Maintainer adjudication remains pending. Selected visible policy passages, captured 11 November 2026. Excerpt, not a full page or raw HTTP response.
- https://www.springernature.com/gp/policies/editorial-policies/ai-editor-guidance
Verified for this PR. Maintainer adjudication remains pending. Selected visible policy passages, captured 11 November 2026. Excerpt, not a full page or raw HTTP response.
Also held, but not the one to cite
- https://group.springernature.com/de/group/media/press-releases/sn-policy-responsible-ai-use/53209516
Official announcement. The framework predates the announcement in the archive evidence held here.
- https://www.springernature.com/gp/policies/editorial-policies/using-ai-in-research
Verified for this PR. Maintainer adjudication remains pending. Selected visible policy passages, captured 11 November 2026. Excerpt, not a full page or raw HTTP response.
What it says, word for word
Use of AI by peer reviewers
Delegating peer review to an LLMSpringer Nature · nature-portfolio-ai · checked against the page
Describe how AI was used, confirm independent judgement and take accountability for all content in reportSpringer Nature · peer-review-ai · checked against a saved browser extract
AI may support communication and exploration under human oversight. Reviewers retain independent expert judgement and must not delegate critique or disclose manuscript content to unsecured or public tools. The reviewer guidance specifies declaration in the report.
The publisher's or editor's own AI use on submitted manuscripts
Nature Portfolio journals occasionally use artificial intelligence tools developed or approved by Springer Nature to support different tasks in the generation of accessory content or to support the editorial process.Springer Nature · nature-portfolio-ai · checked against the page
Any substantive use of artificial intelligence beyond the above will be declared on an individual article basis.Springer Nature · nature-portfolio-ai · checked against the page
Ancillary support is described in the policy; substantive use beyond those listed activities is declared per article.
Translation of manuscript text for the purpose of publishing
formatting; translation; summarising author-written textSpringer Nature · manuscript-preparation · checked against a saved browser extract
Clearly describe use and confirm author accountabilitySpringer Nature · manuscript-preparation · checked against a saved browser extract
Author guidance places translation in green with a declaration. This is more explicit than the umbrella green-tier wording.
Writing or drafting manuscript content
Generating outlines; drafting sections based on author input;Springer Nature · manuscript-preparation · checked against a saved browser extract
Clearly describe use and confirm author accountabilitySpringer Nature · manuscript-preparation · checked against a saved browser extract
Drafting from author input is amber. Disclosure does not authorize replacement of scientific judgment.
Generation, refinement, correction, editing or formatting of images, diagrams or other figures for illustrative purposes only
creating visual content with verifiable source data or materialSpringer Nature · manuscript-preparation · checked against a saved browser extract
Clearly describe use and confirm author accountabilitySpringer Nature · manuscript-preparation · checked against a saved browser extract
Grounded visual content is amber and must be labelled. Ungrounded generation is not covered by this permission.
AI used in the research process itself — analysis, code execution, agentic workflows
AI may support your work, but you remain fully accountable for all research decisions, outputs, and conclusions.Springer Nature · research-use · checked against a saved browser extract
Research support under human accountability; the captured principle is not a task-by-task permission or a disclosure exemption.
Whether AI-text or integrity screening is run, and whether authors are told
The accuracy and usefulness of such tools is far from definitive, and no tool is able to confirm whether the AI use was safe and policy compliant.Springer Nature · editor-guidance · checked against a saved browser extract
Detector use calls for extreme caution and confidentiality checks. This guidance is not a statement that routine screening occurs.
Threshold for review, notification of the author, and route of appeal
The presence of an AI declaration does not indicate concern. Editors may seek clarification where: A declaration lacks sufficient detail.Springer Nature · editor-guidance · checked against a saved browser extract
Explains when to seek clarification. Does not create a universal AI-specific appeal right.
STM Association normative body classification only checked 27d ago
Source of the nine manuscript-preparation activities used here. A shared vocabulary is not permission: task definitions are not counted as publisher rules. The retained recommendations are advisory.
Cite this
What it says, word for word
AI used in the research process itself — analysis, code execution, agentic workflows
This classification addresses only the use of AI assistance for the preparation of manuscripts intended for publicationwhat is said instead · STM Association · classification-2025 · checked against the page
Classify uses of AI in research processes.what is said instead · STM Association · classification-2025 · checked against the page
Research-process AI is explicitly outside this manuscript-preparation classification.
Refinement, correction, editing or formatting the manuscript to improve clarity of language
2023 STM Guidelines recommend that disclosure is not necessary for this use of automated tools.STM Association · classification-2025 · checked against the page
One of only two of the nine activities carrying an inherited permission level.
Presentation of any kind of content generated by AI tools as though it were original research data/results from non-machine sources
2023 STM Guidelines recommend prohibiting this use of AI tools.STM Association · classification-2025 · checked against the page
Taylor & Francis publisher rule with exemptions checked 0d ago
Task-specific rules verified on 11 November 2026. Unassessed activities remain blank.
Individual journals genuinely differ from this policy — check your specific journal as well.
Cite this
- https://taylorandfrancis.com/our-policies/ai-policy/
Verified for this PR. Maintainer adjudication remains pending.
What it says, word for word
Translation of manuscript text for the purpose of publishing
Copyediting, language improvement and translationsTaylor & Francis · ai-policy · checked against the page
Name tool/version, how and why. Individual journals can restrict use beyond language improvement.
Generation, refinement, correction, editing or formatting of images, diagrams or other figures for illustrative purposes only
Taylor & Francis does permit use of Generative AI tools to assist research data visualisations and conceptual illustrations of processes, flow diagrams or teaching illustrations.Taylor & Francis · ai-policy · checked against the page
Does not permit creation or manipulation of research or clinical-testing outputs.
Use of AI by peer reviewers
However, Generative AI may be utilised to assist with improving review language.Taylor & Francis · ai-policy · checked against the page
No manuscript upload, substantive analysis, summarization or generated review reports.
World Association of Medical Editors normative body rule with exemptions checked 0d ago
Recommendations revised 31 May 2023. Research uses are reported in Abstract and Methods; drafting in acknowledgments. These are recommendations, not every journal’s rule.
Cite this
- https://wame.org/pdf/Chatbots-Generative-AI-and-Scholarly-Manuscripts.pdf
Verified for this PR. Maintainer adjudication remains pending.
What it says, word for word
Prompts, model versions, session records retained as evidence
In the interests of enabling scientific scrutiny, including replication and identifying falsification, the full prompt used to generate the research results, the time and date of query, and the AI tool used and its version, should be provided.World Association of Medical Editors · ai-guidance · checked against the page
Recommendations revised 31 May 2023. Research uses are reported in Abstract and Methods; drafting in acknowledgments. These are recommendations, not every journal’s rule.
Wellcome funder rule with exemptions checked 0d ago
Grant applications, not journal articles. Minimal language, translation and formatting uses are exempt.
Cite this
- https://wellcome.org/research-funding/guidance/policies-grant-conditions/use-of-generative-ai
Verified for this PR. Maintainer adjudication remains pending.
What it says, word for word
Writing or drafting manuscript content
Applicants for Wellcome funding can use generative AI tools when preparing a grant application but must tell us about any substantive use of generative AI on their application form.Wellcome · ai-guidance · checked against the page
Grant applications, not journal articles. Minimal language, translation and formatting uses are exempt.
Wiley publisher per task matrix checked 27d ago
Purpose-specific guidance covers research-process use, agentic workflows and disclosure. Its own policy-development disclosure is recorded separately from author requirements.
Cite this
- https://www.wiley.com/publishing/article/ai-guidelines
Redirect verified on 15 October 2026 from https://www.wiley.com/en-us/publish/article/ai-guidelines/. Policy-body wording is unchanged; the changed hash reflects site navigation and formatting.
Also held, but not the one to cite
- https://authors.wiley.com/ethics-guidelines/index.html
Recorded until 2026-09-11 as https://authorservices.wiley.com/ethics-guidelines/index.html, which now redirects here.
What it says, word for word
Writing or drafting manuscript content
AI use in manuscript development is not grounds for rejection.Wiley · ai-guidelines · checked against the page
Disclosure is routed by purpose — drafting and editing to Acknowledgements.
Translation of manuscript text for the purpose of publishing
In your Acknowledgments when AI assists with manuscript drafting, editing, translation, or formatting.Wiley · ai-guidelines · checked against the page
AI used in the research process itself — analysis, code execution, agentic workflows
You may have heard these referred to as: agentic AI, autonomous agents, or agent workflowsWiley · ai-guidelines · checked against the page
a goal and allowing it to carry out multiple steps independentlyWiley · ai-guidelines · checked against the page
Agentic search: AI independently generates or executes search strategiesWiley · ai-guidelines · checked against the page
Names agentic workflows and asks for the audit trail — prompts, parameters, settings and a code repository.
Use of AI by peer reviewers
We require disclosure of AI when it's used for drafting assistance in your reviews.Wiley · ai-guidelines · checked against the page
Editors have the right to disregard peer reviewer reports that show signs of inappropriate AI use.Wiley · ethics-guidelines · checked against the page
Organising, prioritising, tone and translation are permitted; uploading the manuscript is not.
The publisher's or editor's own AI use on submitted manuscripts
Authors have not given permission for reviewers or editors to share their unpublished work with AI Technologieswhat is said instead · Wiley · ai-guidelines · checked against the page
Wiley's author-facing guidance does not state what AI Wiley itself runs on a submission. The quote below is the closest it comes, and it is a restriction on reviewers rather than a statement of Wiley's own use.
Whether AI-text or integrity screening is run, and whether authors are told
Wiley routinely screens submitted manuscripts for duplicated text using tools such as Crossref Similarity Check.what is said instead · Wiley · ethics-guidelines · checked against the page
Wiley discloses duplicate-text screening and says nothing about AI-text detection, despite having announced generative-AI content detection in 2024.
Threshold for review, notification of the author, and route of appeal
The final decision on whether using AI Technology is appropriate or permissible for a submitted manuscript or a published article belongs to the journal's editorwhat is said instead · Wiley · ethics-guidelines · checked against the page
No threshold for triggering review, no duty to notify the author, and no route of appeal is stated. The quote records what is granted instead: editorial discretion.
Prompts, model versions, session records retained as evidence
The final draft was revised by the author using Claude 4 Sonnet and GPT-5 (August 2025) and then polished by a human copyeditor with expertise in research communication.Wiley · ai-guidelines · checked against the page
Wiley discloses that this very guidance document was AI-drafted — specific, in the register of methods, attached to the artefact, with no apology. The norm, demonstrated.
Method
What this does and does not claim
Word for word, or not at all
Every verdict carries a short quote, its source, and how it was obtained. All 140 quotes on this page are checked automatically against the stored copy of the page they came from; the build fails if any quote cannot be found. An earlier version of this site carried a sentence in quotation marks that turned out to be a research assistant’s paraphrase. That is what this check exists to prevent.
Published policy only
This records what organisations have published. It cannot see what editors actually do. Across seventeen cardiovascular journals studied, fifteen ran AI-detection software and one said so publicly — the gap between stated and practised is real, and this site does not close it.
Coverage, stated plainly
36 organisations. Coverage is recorded activity by activity. If your journal, funder or institution has no entry, that means it has not been assessed — not that it is silent.
Where the evidence comes from
38 of 53 sources are fetched directly. 5 retain dated Internet Archive captures; their current pages are checked separately and carry a live-check date. 10 uses a saved excerpt from the rendered browser page, labelled as such. A fixed archive capture does not establish that the live policy is unchanged. Missing live checks enter the monitoring queue.
Contradictions, counted honestly
0 substantive internal contradictions and 4 dating or tagging oddities across 36 organisations. A stale HTML title tag is not a policy contradiction and is not counted as one. Differences in audience or product scope are kept separate.
0 cells still await a quote
Verdicts recorded without captured wording are marked and published, never asserted as verified. Where we could not find wording at all, the activity remains unassessed rather than being labelled silent.
This page declares itself unmaintained automatically once its oldest record passes 550 days without re-checking. The closest precedent to this project still serves a live page, with no staleness warning, four years after its last update.
Corrections and right of reply
If you work for an organisation recorded here and think a record is wrong, we want to fix it. Every record names its source and quotes the wording it relies on, so corrections can be specific. Open an issue and we will respond. Corrected records carry the date they changed, and the change stays in the public history.