03 · What You Need to Know
How Journal Notification Requirements Differ Across AI Activities
Disclosure and Notification Are Related but Different Obligations
AI disclosure ordinarily means providing information about an AI tool's contribution through a manuscript statement, methods description, submission declaration, acknowledgment, or another location required by the journal.
Journal notification is a communication to the editor or editorial office about an activity or concern. It may be necessary because a required disclosure was omitted, because a submitted declaration is inaccurate, or because a previously unrecognized problem affects the manuscript.
These obligations overlap, but they are not identical. An author may be required to describe AI-assisted analysis in the methods section without sending a separate explanatory email when the original submission already contains all required information. Conversely, a researcher who discovers a consequential error in AI-generated code may need to notify the editor even if the AI tool was properly disclosed.
Consequently, the decision should consider both reporting compliance and the integrity of the research.
Basic Language Corrections May Be Exempt
Elsevier's journal policy, updated in June 2026, distinguishes basic spelling, grammar, and punctuation checks from more substantive AI-assisted changes. Under that policy, the former do not require an AI declaration, whereas substantive changes to sentence structure, organization, or content require disclosure.
This distinction is important because contemporary writing software often incorporates AI functionality into ordinary editing operations. Merely activating an AI-enabled spelling checker does not necessarily mean that the author has performed an activity requiring a declaration.
However, the exception should not be generalized across every journal. ICMJE guidance calls for disclosure of AI-assisted technologies used in producing submitted work and identifies writing assistance as an activity to report. Journal-specific instructions remain decisive.
Researchers should also distinguish simple proofreading from AI-generated rewriting. A system that corrects a typographical error performs a different function from one that substantially reconstructs the argument or introduces new claims.
Substantive Writing Assistance Often Requires Disclosure
Generative AI can rewrite paragraphs, reorganize discussions, suggest arguments, draft abstracts, and produce literature summaries. Such activities may influence the intellectual presentation of a study even when the researcher subsequently edits the output.
Under Elsevier's current journal guidance, substantive AI-assisted manuscript preparation is subject to declaration requirements. The statement should identify the tool and purpose of use, with authors retaining responsibility for the content.
Springer Nature and other publishers maintain their own policies. Researchers should consult the particular journal's author instructions rather than assuming that one publisher's requirements govern the entire scholarly publishing system.
If a required declaration was omitted, notifying the editorial office allows the journal to determine how the submission or publication record should be amended.
AI Used in Research Methods Is Different From AI Used to Polish Prose
AI may participate directly in research design, data collection, transcription, coding, statistical analysis, image processing, or interpretation. These activities can affect the production of evidence rather than merely its presentation.
ICMJE recommends describing AI use for data collection, analysis, or figure generation in the methods section. Elsevier similarly distinguishes AI-assisted manuscript preparation from AI tools used as part of the research process, which should be documented in the methodology.
Failure to report substantive AI involvement may therefore leave the methodological account incomplete, even where the numerical results or conclusions are correct.
The relevant question is whether readers and reviewers have received the information necessary to understand how the study was conducted. This is closely connected to how the seriousness of undisclosed AI use depends on the task performed.
A Practical Classification of AI Activities
| AI activity |
Typical disclosure consideration |
Notification implication |
| Basic spelling or punctuation checking |
May be expressly exempt under a journal's policy |
Usually unnecessary where an applicable exemption clearly covers the activity and no other concern exists |
| Substantive rewriting or paragraph generation |
Often reportable as AI-assisted manuscript preparation |
Contact the journal if a required declaration was omitted |
| Literature synthesis or generation of scholarly claims |
May require disclosure and independent source verification |
Notify when reporting requirements were missed or unsupported claims affect the manuscript |
| AI-generated statistical or analytical code |
May require methodological documentation depending on its role and the applicable policy |
Notify if a required description is missing or the results may be unreliable |
| AI-assisted qualitative coding or classification |
Substantive methodological involvement generally warrants transparent description |
Seek editorial guidance if omitted from a submitted or published methodological account |
| AI-generated or altered research images |
Subject to potentially strict publisher, journal, and disciplinary rules |
Prompt notification is appropriate if the activity breached policy or compromises the evidence |
This classification is a decision aid, not a universal journal policy. The same tool may perform different functions, and disclosure requirements can change according to how its output was used.
Does the Manuscript's Publication Stage Change the Response?
Yes. The appropriate communication and corrective mechanism may differ according to whether the work is being prepared, reviewed, accepted, or already published.
Before submission.
Add any required disclosure or methodological description before submitting. No separate post-submission notification is necessary because the journal has not yet received the manuscript.
During editorial review.
Contact the editorial office if a required declaration was omitted or an earlier response to a submission question was inaccurate. Follow the journal's instructions for submitting revised information.
After acceptance but before publication.
Notify the production or editorial team promptly so the required information can potentially be incorporated before the version of record is released.
After publication.
Ask the journal to assess whether the omission warrants a formal amendment, correction, or other action under its post-publication procedures.
The later an omission is identified, the more important it becomes to distinguish an ordinary manuscript revision from a change to the public scholarly record.
When Does Notification Become Especially Important?
Several circumstances strengthen the case for prompt editorial communication, even if the journal's AI disclosure wording is ambiguous.
These include AI-generated references that may not exist, unverified analytical outputs, inaccurate descriptions of research methods, undisclosed image alterations, or AI involvement that may have compromised confidential information.
Notification may also be necessary where authors discover that they answered a direct submission question inaccurately. An inaccurate declaration should not be treated as inconsequential merely because the omitted activity appears relatively minor.
Watch Out
A disclosure exemption is not permission to submit inaccurate findings, fabricated references, manipulated evidence, or confidential research information to an unauthorized AI service. These problems may require action independently of whether the AI activity itself was reportable.
Should Researchers Notify the Journal When the Policy Is Ambiguous?
Not every uncertainty requires a formal incident report. Researchers can first examine the author instructions, editorial policies, submission declarations, and any relevant policy FAQs.
If the activity was clearly exempt and the manuscript remains accurate, further communication may add little value. If the activity was substantive and the reporting requirements are unclear, a concise request for clarification is reasonable.
Describe the actual activity rather than asking whether "AI use" generally requires disclosure. Editors can assess a specific account of what a tool did more effectively than an abstract question about artificial intelligence.
Where the omission could materially affect the reliability or transparency of published work, researchers should not delay necessary notification while debating fine distinctions in disclosure terminology.
What Should a Notification Contain?
A useful notification identifies the manuscript or article, describes the AI tool and its role, explains what was previously declared, and states why the authors believe clarification or correction may be necessary.
If the article has been published, include its DOI or other publication identifier. Where verification has already been completed, briefly describe the checks and their outcomes. If the effect on the results remains uncertain, say so explicitly.
The communication should request the journal's preferred procedure without asserting that a correction, retraction, or finding of misconduct is inevitable.
Keep a record of the correspondence and any subsequent editorial decision.
04 · A Practical Example
Three Researchers Discover Different Undisclosed AI Activities
Hypothetical Example
One Journal, Three Different AI Contributions
Three authors independently discover previously undisclosed AI use in articles published by a journal that follows a policy expressly exempting basic spelling, grammar, and punctuation checks but requiring disclosure of substantive AI-assisted writing and methodological use.
Researcher A used an AI-enabled spelling checker to correct typographical errors. No substantive text was generated, and the activity falls within the journal's stated exemption.
Researcher B used a chatbot to rewrite several discussion paragraphs. The author verified the scientific claims, but the journal's required AI declaration was omitted.
Researcher C used generative AI to produce statistical code for the main analysis. The published methods section does not identify this contribution, and the code has not been independently checked.
| Researcher |
Assessment |
Appropriate next action |
| A |
Activity is covered by an explicit disclosure exemption |
Document the assessment; no journal notification is ordinarily needed |
| B |
Required manuscript-preparation declaration was omitted |
Contact the journal about adding the missing disclosure |
| C |
Methodological reporting is incomplete and analytical reliability remains uncertain |
Preserve the code, independently verify the analysis, and promptly inform the editor of the material concern |
All three researchers used AI, but identical notification decisions would overlook meaningful differences. The journal's policy, the function performed, and the consequences of the omission determine the response.
06 · What This Means for You
Use the Journal's Requirements and the AI Contribution to Decide Whether to Notify
The most useful approach is to make a documented, task-specific determination rather than adopting a blanket rule that every AI interaction must be reported or that only major AI involvement matters.
First, identify the applicable disclosure requirement. Then establish the AI activity and compare it against that requirement. Finally, determine whether any separate concern about accuracy, research integrity, confidentiality, or reproducibility requires editorial attention.
A simple decision framework
If the activity clearly falls within an explicit disclosure exemption
Retain a record of the policy and assessment; notification is generally unnecessary unless another concern exists.
If the activity was reportable but omitted
Contact the journal and provide the information needed to amend the record.
If the policy is ambiguous and AI made a substantive contribution
Request editorial clarification with an accurate description of the activity.
If AI may have compromised the findings, ethics, or confidentiality
Address the substantive concern promptly and notify the appropriate editorial or institutional authority.
When a required disclosure has already been omitted from a published article, the question becomes how the publication record can be corrected. If the activity also raises questions about scientific honesty or reliability, assess whether the undisclosed AI use constitutes a research-integrity concern.
Where an omission was discovered long after publication, also consider whether the relevant policy existed at the time. The distinction between an honest misunderstanding and deliberate concealment may matter for institutional evaluation, although the published record may require amendment in either circumstance.
