Journal AI Policies for Authors in 2026

You used AI to polish a paragraph, translate your abstract, or suggest references. Before you submit, what should you tell the journal?

There is no single rule covering every task. A journal may exempt limited language editing, require disclosure of translation, and prohibit generating particular content. Even within one publisher, different policy pages or journals may use different definitions.

To read these policies, start with three questions: What did the tool do? Is that use permitted, and does it require disclosure? Where should you report it, and how much detail is needed? This guide explains that framework, then applies it to selected official policies and an abstract-translation example.

Policy sources checked on October 11, 2026. This guide concerns journal submissions. Always check the target journal’s current Instructions for Authors and the publisher policy it references.

Start with five categories of AI use

A tool’s name does not tell an editor how you used it. The same service can edit language, generate a paragraph, suggest citations, or write code. Classify the activity rather than treating everything as ‘AI assistance.’

Category	What it covers	Examples
Language Editing	Improving existing author-written text while preserving its meaning and claims	Grammar correction, wording suggestions, consistency checks, and style refinement
Writing/Drafting	Generating new text, explanations, or formulations of ideas for the manuscript	A preliminary abstract, an introduction paragraph, or a new explanation of a finding
Images/Figures	Creating or modifying visual material included in the submission	Conceptual illustrations, experimental schematics, or changes to a research image
Data/Visuals	Assisting with data processing, analysis, or visual presentation of data	Analysis code, data cleaning, plots, charts, and other data visualizations
References	Assisting with source discovery, selection, citation, or bibliography preparation	Suggested papers, bibliographic checks, citation formatting, and reference organization

These are reading categories, not universal policy definitions. A single task can belong to more than one category. For example, AI-generated plotting code involves Data/Visuals and produces a figure, so both methodological reporting and figure rules may matter.

Translation needs an explicit check even when you are translating your own text. Do not assume that a language-editing exemption covers it. Likewise, a prompt to ‘improve the English’ may produce extensive restructuring or new explanations. Examine what you actually retained.

Read the permission and disclosure labels correctly

After identifying the activity, ask two separate questions: is it allowed, and must it be reported? The following labels make a comparison table easier to read.

Policy label	Meaning	What the author should do
No disclosure required	The policy expressly exempts the specified activity from disclosure	Confirm that your use stays within the exemption; retain a record of what you did
Disclosure required	The specified use is permitted under stated conditions and must be reported	Identify the tool, activity, scope, and other details the journal requests
Prior permission required	The specified use needs approval before the relevant submission or use	Contact the editor or publisher and retain the approval
Prohibited	The policy does not permit the specified activity	Do not treat a disclosure statement as permission to use the prohibited material
Not addressed	The reviewed policy does not explicitly resolve this particular activity	Check linked guidance and ask the editorial office if the answer remains unclear

A ‘not addressed’ entry does not mean ‘allowed’ or ‘no disclosure required.’ It identifies a gap in the guidance reviewed. A ‘disclosure required’ entry also does not mean every use in that category is acceptable: restrictions and exceptions still apply.

Conditional permission is an additional qualification. Read the condition beside the label, such as permission for explanatory illustrations but not evidential images. Approval, purpose, verification, and licensing conditions can materially change the answer.

Find the right place to disclose the use

Disclosure is not always a sentence in Acknowledgments. Its location can depend on whether AI helped communicate the research, contributed to the method, or produced a visual.

Reporting location	Purpose	What to check
Acknowledgments	Describing assistance with preparing the manuscript	Required wording, tool identity, affected sections, and author review
Methods	Explaining AI’s contribution to the research process	Enough technical and validation detail to understand and reproduce the work
Separate AI declaration	Giving a distinct account of manuscript-preparation assistance	The prescribed heading, position, and statement format
Cover letter or submission form	Providing information directly to the editorial team	Whether this is required in addition to disclosure within the manuscript
Supplementary material	Providing supporting records or detailed technical information	Whether prompts, outputs, settings, or other records must be supplied
Figure caption or legend	Identifying AI’s role in a particular visual	Whether a caption disclosure and a general or Methods disclosure are both needed

For example, Elsevier uses a separate declaration before the references for manuscript preparation and Methods for research use. IEEE places disclosure of AI-generated article content in Acknowledgments. Follow the journal’s own route rather than assuming these locations are interchangeable.

Understand three levels of disclosure detail

A brief acknowledgment and a complete prompt log provide different information. It helps to distinguish three levels of detail when reading a journal’s requirements. These are a practical framework, not a universal reporting standard or a choice authors can make regardless of the journal’s instructions.

Level 1 Tool identity and the fact of use

This identifies the tool and states the activity. It can serve as a starting point, but may leave the extent of use and review unclear.

The authors used [tool name] for language editing of the manuscript.

That sentence alone may not meet the journal’s requirements. It does not explain which sections were affected, what changes were made, or how the output was checked.

Level 2 Tool details purpose scope and author review

This explains what the tool contributed and how the authors handled it. Include a model or version and access date when required or relevant, using only details you can verify.

The authors used [tool name and model/version, accessed date] to suggest language revisions to the author-written Abstract and Introduction. The authors compared the suggestions with the original text, revised them where needed, and approved the final wording. The authors take responsibility for the final content.

This original example is adaptable wording, not an official publisher template. Describe only the review steps that actually occurred. If your journal supplies required wording, follow that wording.

Level 3 Prompts outputs and supporting records

More detailed reporting can include the full prompts, generated outputs, model settings, affected sections, and validation records. These may belong in Methods, an appendix, or supplementary material, depending on the study and journal.

Keep such records when they could be needed to explain or reproduce the work. Do not assume every language-editing use requires publishing all prompts, or that a short acknowledgment is sufficient when AI contributes to a research method.

Compare selected journal and publisher policies

Compare these selected policies across the five activity categories. Each row links to official guidance. Read each condition alongside the disclosure requirement; a publisher-wide summary cannot replace the target journal’s instructions.

Cell format: permission or disclosure status; reporting location; required detail. “Declaration” means a separate AI-use statement. L2 denotes use-specific information beyond tool identity; the fields shown are the ones summarized for that policy, not a universal checklist. L3 denotes prompts, outputs, or supporting records, including records requested rather than routinely published. “Not specified” means this reviewed guidance does not prescribe a level.

Publisher and sources	Language Editing	Writing/Drafting	Images/Figures	Data/Visuals	References
Elsevier Official policy	Basic checks: no disclosure. Substantive changes: disclose. Declaration; L2: tool, purpose, review.	Supportive use: disclose; preserve author contribution. Declaration; L2: tool, purpose, review.	Explanatory visuals: conditional. Caption + declaration; L2: tool/version, use. Primary-image generation prohibited; research-method exception.	Actual-data visualization: conditional. Methods; L2: tool/version, developer, reproducible use. Fabrication prohibited.	Basic reference managers: exempt. AI source assistance: disclose. Declaration; L2: tool, purpose, review. Verify sources.
Nature Portfolio Official policy Journal example	Defined copyediting: no disclosure. Autonomous content creation falls outside exemption.	Other LLM use: disclose. Methods or suitable alternative. Detail level not specified.	Generative images generally prohibited; limited exceptions. Research-method use: Methods. Detail level depends on exception.	Research-method AI use: disclose in Methods. Non-generative image enhancement: caption. Detail level not specified here.	LLM use: general Methods disclosure applies. No separate reference-specific level stated. Verify every source.
Springer Official journal policy Publisher guidance	Low-risk language polishing: permitted; disclosure encouraged. Location and detail: check journal.	Extensive writing assistance: risk-dependent; disclose and verify. Location and detail: check journal.	Risk-dependent; deceptive deepfakes prohibited. No category-wide location or detail level prescribed.	Cleaning: low-risk. Interpretive analysis: disclose and verify. Fabrication prohibited. Location/detail: check journal.	Fabricated citations prohibited. Other assistance: apply risk framework and verify. Location/detail: check journal.
Wiley Official ethics guidance Disclosure guidance	Sole spelling/grammar/general editing: exempt. Substantial edits: disclose in Acknowledgments. Detail: follow declaration guidance.	Development/translation: disclose. Acknowledgments for preparation; Methods for research. Detail: follow declaration guidance.	Conditional; editorial assessment applies. Caption/legend disclosure. No single detail level for all visuals.	Research/data assistance: disclose in Methods. Describe purpose and assess influence on findings. L2-type use details.	Independent AI literature work: disclose in Methods. Routine formatting/organization: exempt. Verify sources; detail depends on use.
Taylor & Francis Publisher guidance Current author guide	Generative refinement: permitted; disclose. AI usage statement; L2: tool/version, use, author responsibility.	First drafts of manuscripts/sections prohibited. Specified supportive uses: disclose. AI usage statement; L2.	Illustrative diagrams: conditional. Legend + AI statement; L2: tool/version, use. Research/clinical image manipulation prohibited.	Verified coding/visualization: conditional. Methods + AI statement; L2. L3 records retained for request; manipulation prohibited.	Verified suggestions permitted; entire bibliography generation prohibited. AI statement; Methods for literature review. L3 records on request.
Sage Official policy	Assistive suggestions to own text: no disclosure. Generative contributions outside this exemption: disclose.	Permitted generative contributions: disclose at submission. Methods or Acknowledgements. Detail level not specified.	Representative illustrations: conditional; disclose in Methods/Acknowledgements. Factual image fabrication prohibited. Detail level not specified.	Visualization/code assistance: conditional; disclose in Methods/Acknowledgements. Separate analysis/data restrictions apply. Detail level not specified.	Literature/reference assistance: disclose in Methods/Acknowledgements. Fabricated references prohibited. Detail level not specified.
IEEE Official policy	Editing/grammar enhancement: disclosure generally optional, recommended. Generated content: use general disclosure rule.	Generated text: disclose. Acknowledgments; L2: system, sections, extent.	Generated figures/images: disclose. Acknowledgments; L2: system, sections, extent. Other integrity rules apply.	Generated code/content: disclose. Acknowledgments; L2: system, sections, extent. Research reporting still applies.	No separate reference-specific rule stated here. Generated article content: Acknowledgments; L2. Verify sources.
ACS Publications Official policy	All AI use: disclose in submission. Location/detail depend on use; text generation follows Acknowledgments rule.	Generated text: disclose in Acknowledgments. L2: when and how. Substantial use: fuller Methods/appropriate section details.	Generated images: Acknowledgments + caption. L2: when/how; cover tool details. AI-generated ToC graphics prohibited.	All AI use: disclose in submission. Substantial use: Methods/appropriate section. No single category-wide detail level.	All AI use: disclose in submission. No separate reference-specific location or detail level. Verify sources.
Emerald Publisher guidance	Generative copyediting: disclose. Methods and Acknowledgements, or appropriate section. L2: tool/version, modified content. Standard non-generative checks exempt.	New submission text prohibited. Scholarly critique examples: exception; flag/cite. Translation of own work: declare; L2.	Illustrative only: conditional. Label/caption + general declaration; L2: tool/version, content. Evidential fabrication and AI cover art prohibited.	Existing-data visualization: conditional. Methods/Acknowledgements; L2: tool/version, content. Generated/manipulated data/results prohibited.	Existing-reference editing: verify and disclose AI use; L2. General declaration location. Bibliography creation by AI prohibited.

Sources checked on October 11, 2026. Levels 1–3 are our analytical framework, not official publisher ratings. An unspecified detail level does not remove the obligation to explain AI use adequately. Several rows summarize general AI rules where a category-specific rule is absent.

Check the exact Springer Nature policy linked by your journal. The Nature Portfolio example exempts defined copyediting, while Springer’s journal policy uses a risk framework and group guidance calls for declaration. Seek clarification if these leave the requirement unclear.

Apply the framework to an AI translated abstract

Suppose you wrote an abstract in Korean, used AI to translate it into English, and revised the result before submission. Here is how to check that workflow.

·       Classify the activity. Record translation of author-written text, plus any substantial rewriting or new content you retained. Do not reduce the description to grammar checking.

·       Check permission and disclosure. Read the target journal’s translation rules. Elsevier, for example, explicitly requires disclosure of AI translations included in the manuscript.

·       Identify the reporting location. Determine whether the journal requires Acknowledgments, a separate declaration, a submission response, or more than one location.

·       Match the detail to the requirement. Record the tool, relevant model/version and date, the translated section, and the review actually performed.

·       Verify the result. Have someone who understands both languages and the research compare terminology, numbers, units, negation, and claim strength with the original.

For this hypothetical workflow, a statement could read:

The author-written Abstract was translated from Korean into English using [tool name and model/version, accessed date]. [Reviewer name or initials], who is proficient in both languages and familiar with the research, compared the translation with the Korean original and checked its scientific meaning. The authors revised and approved the English Abstract and take responsibility for its content.

Use this account only if that review occurred, and adapt it to the journal’s required format. Taylor & Francis also specifies the responsible author and acknowledgment of other translation-checking contributors where applicable.

Check references and visuals on their own terms

A writing disclosure does not settle every AI-related requirement. For references, distinguish routine formatting from AI-assisted source selection or generation. Verify that each source exists, that its bibliographic details are correct, and that the original supports your claim.

For visuals, identify what the image represents. Conceptual diagrams, data-derived plots, and microscopy images serve different purposes. Check the applicable rule for each visual; permission for one does not extend automatically to another.

Also check the tool’s terms before uploading unpublished text or data. Review retention, model-training use, confidentiality, and publication rights, alongside any institutional or participant-data restrictions.

Prepare your disclosure and final manuscript

Once you have checked the policy, the next article in this series, AI Disclosure Statements for Research Papers With Examples, helps you turn your actual workflow into a clear statement.

For a final language review, Wordvice Academic Editing can help improve clarity, grammar, consistency, and academic style. Authors remain responsible for the research, the account of AI use, and compliance with the target journal’s requirements.