03 · What You Need to Know
How to Interpret Unclear AI Disclosure Requirements Without Misrepresenting the Research
Why Are Journal AI Disclosure Policies Sometimes Ambiguous?
Generative AI technologies perform activities that overlap with several established research practices.
A language model may function as a grammar editor, programming assistant, literature-search aid, data classifier, or preliminary analytical consultant.
Journal policies may describe AI-assisted writing without explicitly addressing these other functions.
Some instructions also use broad expressions such as "significant assistance," "substantial contribution," or "AI-generated content" without defining their boundaries.
These terms can be difficult to apply when a researcher uses AI to generate code but writes the manuscript independently.
Ambiguity may also arise because publisher-level guidance and individual journal instructions were developed or revised at different times.
The researcher should therefore identify precisely which activity creates uncertainty rather than treating the entire AI policy as unclear.
Does the Absence of an AI Policy Mean AI Use Is Permitted?
Not necessarily. The absence of an explicit AI policy does not establish that every form of AI assistance is authorized.
Other requirements may still apply, including research ethics approval, confidentiality obligations, data-protection rules, and ordinary standards of accurate methodological reporting.
For example, a journal may have no dedicated rule concerning AI-assisted interview coding, but the researcher must still explain how the qualitative analysis was conducted.
Likewise, a research ethics protocol may restrict uploading participant transcripts to external services regardless of whether the journal mentions AI.
However, policy silence should not automatically be interpreted as a prohibition either.
The appropriate conclusion is that the specific requirement remains unresolved and may need clarification.
Does an Ambiguous Policy Mean Researchers Should Disclose Everything?
Not automatically. Reporting every interaction with AI may produce lengthy declarations containing little information relevant to the research.
For example, a researcher might use AI repeatedly to correct punctuation while also using it once to generate a statistical procedure that materially influences the findings.
The second activity may be more consequential despite involving fewer interactions.
Researchers should therefore distinguish routine assistance from activities that affect manuscript substance or research methodology.
Where the journal's policy requires broad disclosure, that requirement should be followed.
Where the policy remains unclear, researchers should describe the actual activities and seek clarification rather than assume that the number of prompts determines what must be reported.
Which Policy Should Researchers Check First?
The intended journal's current author instructions are the immediate starting point for manuscript submission.
Researchers should then examine the publisher's general AI policy and any additional submission requirements.
Institutional, funding, contractual, and research ethics obligations may also apply independently.
| Policy Source |
What to Examine |
Why It Matters |
| Journal author guidelines |
Permitted AI use, disclosure categories, statement placement, and required wording. |
Establishes instructions for the intended submission. |
| Publisher AI policy |
General rules, exceptions, restrictions, and definitions. |
May clarify or supplement journal-level instructions. |
| Submission system |
AI-use questions, declarations, and author confirmations. |
May impose reporting requirements not repeated in the manuscript template. |
| Institutional policy |
Research integrity, data security, and permitted AI assistance. |
May impose independent obligations. |
| Research ethics approval |
Approved data handling, participant protections, and methodological procedures. |
May restrict AI processing of research materials. |
| Funding or research agreement |
Confidentiality, reporting, and contractual conditions. |
May establish additional requirements. |
These sources should be read together, but they should not be assumed to have identical scope or authority.
What If the Journal and Publisher Policies Appear to Conflict?
Researchers should first determine whether the instructions actually conflict or address different aspects of AI use.
For example, a publisher may specify a general declaration requirement while the journal provides additional instructions concerning methodological reporting.
These requirements may be complementary rather than contradictory.
However, a genuine conflict may arise if one document appears to permit an activity while another appears to prohibit it.
Researchers should not automatically assume that the more recent document or the more permissive statement takes precedence.
The appropriate response is to seek clarification from the journal or publisher using the specific conflicting provisions.
Any clarification received should be retained with the relevant submission records.
What Do ICMJE and COPE Say About AI Disclosure?
The International Committee of Medical Journal Editors (ICMJE) recommends disclosure of AI-assisted technologies and an explanation of how they were used in submitted research.
Its guidance distinguishes AI writing assistance from methodological AI use and emphasizes that human authors remain responsible for the submitted work.
The Committee on Publication Ethics (COPE) similarly emphasizes transparency about AI tools used in manuscript preparation and research activities.
These positions can help researchers understand general publication-integrity expectations.
However, they should not be treated as automatically resolving every journal-specific ambiguity.
A journal may adopt additional requirements or implement the guidance differently.
The researcher should identify the applicable policy rather than assume that one publishing authority's recommendations govern every submission.
How Should Researchers Interpret the Word "Substantial"?
The meaning depends on how the journal defines the term.
In the absence of a definition, researchers should avoid assuming that substantial assistance refers only to the number of AI-generated words.
A single AI-generated analytical procedure may materially affect a study's findings, while numerous grammar corrections may have little effect on its intellectual content.
Substantial assistance may therefore concern the nature and significance of the contribution rather than its volume.
However, this interpretation should not be presented as the journal's official rule unless the policy supports it.
Where the classification affects disclosure, researchers should ask the editorial office how the journal applies the term to the relevant activity.
What If the Journal Mentions AI Writing but Not AI Data Analysis?
The absence of a specific data-analysis provision does not eliminate ordinary methodological reporting responsibilities.
For example, a researcher who uses AI to classify survey responses should describe the coding procedure sufficiently for readers to evaluate how the findings were produced.
The journal's AI-writing declaration may not be the appropriate location for that information.
Researchers should distinguish the obligation to describe methods accurately from the separate requirement to complete a generative AI declaration.
If the journal's reporting instructions remain unclear, editorial clarification may be appropriate.
The question of whether AI use belongs in methods or acknowledgments should be resolved according to its actual role and the journal's instructions.
What If the Journal Mentions ChatGPT but Not Other AI Tools?
Researchers should examine whether the policy regulates a particular application or a broader category of AI-assisted technologies.
A journal may use ChatGPT as an example of generative AI without intending to exclude other language models or related applications.
However, researchers should not automatically extend every provision to technologies with substantially different functions.
For example, a conventional spelling checker, a generative writing assistant, and an automated research classifier may fall under different policy categories.
The relevant issue is the functionality used and the policy's actual wording.
Where uncertainty remains, the researcher should describe the tool and task when seeking clarification.
What If the Policy Does Not Distinguish Grammar Correction From Substantive Writing?
This ambiguity matters because some publishers exempt defined routine AI-assisted copyediting while requiring disclosure of substantive generative assistance.
For example, Springer Nature distinguishes certain copyediting activities from other generative AI uses under its stated conditions.
Other journals may impose broader reporting requirements.
A researcher should therefore examine whether the tool merely corrected existing language or introduced new arguments, interpretations, or content.
If the journal's policy does not clearly classify the activity, the researcher may request clarification.
The question of whether AI grammar correction requires disclosure should not be answered solely by the name of the application used.
What If the Journal Requires Disclosure but Does Not Specify Where?
Researchers should examine the manuscript template, submission checklist, and publisher-level instructions for a designated reporting location.
Possible locations include acknowledgments, methods, a dedicated AI-use declaration, or submission correspondence.
The appropriate location depends on the assistance and the journal's requirements.
For example, AI-generated statistical code may require methodological explanation even if the journal also asks for a separate declaration.
Where the location remains ambiguous, the researcher may ask the editorial office.
The broader guidance on where AI use should be disclosed provides a useful distinction between manuscript preparation and research methods.
Should Researchers Contact the Editorial Office?
Yes, when the ambiguity is consequential and cannot be resolved through the available instructions.
An editorial inquiry is particularly appropriate when uncertainty concerns whether an activity is permitted, whether disclosure is required, or whether the journal expects specific methodological information.
However, not every minor uncertainty requires correspondence.
If the policy clearly addresses the activity, researchers should follow it rather than seek unnecessary confirmation.
When clarification is needed, the inquiry should be concise and specific.
Editors are more likely to provide a useful response when the researcher explains what AI did rather than asking a broad question about whether AI is allowed.
What Should an Editorial Inquiry Include?
A useful inquiry identifies the manuscript context, AI tool, actual assistance, and the specific policy provision creating uncertainty.
It should also ask a question the editorial office can answer directly.
Illustrative Editorial Inquiry
Clarification About AI-Assisted Statistical Code
"I am preparing a manuscript for submission to your journal and would appreciate clarification regarding the generative AI disclosure policy. ChatGPT was used to generate preliminary Python code for statistical analysis. The code was subsequently reviewed, revised, and tested by the research team. Your guidelines address AI-assisted manuscript writing but do not explicitly discuss AI-generated analytical code. Should this assistance be included in the journal's AI-use declaration, the methods section, or both?"
The wording should be adapted to the actual assistance and verification performed.
Researchers should not claim that code was tested or independently validated unless those procedures occurred.
Should Researchers Save the Editorial Response?
Retaining relevant clarification can help document how the disclosure decision was made.
For example, an editorial email may explain whether a particular AI-assisted activity falls within the journal's reporting requirements.
Such correspondence may be useful if questions arise during peer review or later editorial processing.
However, an informal response should not automatically be treated as overriding institutional ethics requirements or other binding obligations.
Researchers should also recognize that journal policies may change between the inquiry and submission.
The response should therefore be considered in its original context and date.
What If the Editorial Office Does Not Respond?
Researchers should review the available policy again and determine whether the submission can proceed without misrepresenting the assistance.
Where AI materially affected the research methodology, accurate methods reporting remains important even if no separate AI declaration requirement is clarified.
A proportionate disclosure may be appropriate when consistent with the journal's instructions.
However, researchers should not assume that silence constitutes approval of an otherwise restricted activity.
If permission is genuinely uncertain, particularly for prohibited or sensitive uses, the researcher may need to obtain clarification before proceeding.
The appropriate decision depends on the nature of the assistance and the consequences of noncompliance.
Should Researchers Use a Voluntary AI Disclosure When the Policy Is Silent?
A voluntary statement may support transparency, particularly when AI materially influenced the research.
However, voluntary disclosure should not be treated as a universal substitute for policy interpretation.
A journal may have specific formatting requirements or restrictions that need to be addressed separately.
Researchers should also avoid disclosing unnecessary confidential information or presenting exploratory interactions as substantive contributions.
Where a voluntary statement is appropriate, it should accurately describe the AI assistance and distinguish writing support from methodological involvement.
The objective is informative transparency rather than indiscriminate reporting.
Does Voluntary Disclosure Make Otherwise Prohibited AI Use Acceptable?
No. Disclosure and permission are separate obligations.
For example, a journal may prohibit certain forms of AI-generated imagery even when authors clearly identify their origin.
Similarly, an ethics approval may restrict uploading participant data to external AI services.
Declaring such activities does not retrospectively authorize them.
Researchers should establish whether the assistance was permitted before deciding how to disclose it.
Where the policy is ambiguous about permission, clarification may be necessary before using the tool or submitting the affected material.
Can Researchers Rely on Another Journal's AI Policy?
Another journal's policy may provide useful context, but it does not govern the intended submission.
For example, a researcher may discover that one publisher exempts routine AI-assisted copyediting.
That exemption does not automatically apply to a journal published by another organization.
Similarly, an AI declaration accepted by one journal may not satisfy another journal's requirements.
Researchers should avoid transferring policy language across publishers without checking its applicability.
The intended journal's instructions and other governing obligations remain decisive.
Should Researchers Document AI Use While Waiting for Clarification?
Yes, particularly when the assistance materially affects the research.
Useful records may include the tool used, purpose, relevant dates, outputs incorporated, and verification procedures.
Such documentation helps researchers prepare an accurate disclosure if the journal later clarifies its requirements.
It may also support methodological transparency and subsequent research-integrity inquiries.
However, not every minor AI interaction requires an elaborate log unless an applicable policy demands it.
The broader question of whether researchers should keep records of AI use becomes particularly relevant when disclosure requirements remain uncertain.
What If the Policy Changes During Peer Review?
Publishers may revise their AI policies while a manuscript is under consideration.
Researchers should follow the journal's instructions concerning which requirements apply to manuscripts already submitted.
If the journal requests an updated declaration, the authors should provide accurate information about the AI assistance actually used.
They should not assume that a new policy automatically applies retrospectively in every circumstance or that the original submission requirements remain unchanged.
Where uncertainty affects compliance, the editorial office should clarify the applicable expectations.
Retaining the original instructions and relevant correspondence can help establish the context of the submission.
What If the Journal's AI Policy Changes After Publication?
A later policy change does not automatically establish that an earlier publication violated the requirements applicable at the time.
However, the original work remains subject to applicable research-integrity and publication standards.
If a substantive omission or error is identified, the authors may need to communicate with the journal regardless of later policy changes.
The appropriate assessment should distinguish the historical requirements, the actual AI assistance, and any continuing obligations concerning the accuracy of the scholarly record.
Researchers should avoid treating a new policy as automatic evidence of past misconduct.
Can Ambiguous AI Disclosure Instructions Create Fairness Concerns?
Yes. Vague requirements may be interpreted differently by authors, editors, reviewers, and institutions.
For example, one researcher may disclose AI-assisted grammar correction while another may reasonably interpret the same activity as exempt.
If a journal has not clearly defined the requirement, inconsistent enforcement can create procedural fairness concerns.
These concerns become more serious when ambiguous disclosure rules are used to support allegations of misconduct.
Investigations should examine the applicable policy, its clarity at the relevant time, the actual conduct, and the available evidence.
Policy ambiguity does not eliminate authors' responsibilities, but it should be considered when assessing whether a particular reporting failure constitutes a violation.
What Should Researchers Do If an Editor Requests More Detail?
Researchers should respond accurately and provide the information requested, subject to applicable confidentiality and ethical restrictions.
For example, an editor may ask which model was used, whether AI-generated code was verified, or how a language model contributed to qualitative coding.
The authors should consult their research records rather than reconstruct details through speculation.
If information was not recorded or is unavailable, that limitation should be stated honestly.
Researchers should not invent model versions, prompts, or verification procedures to satisfy a request for greater specificity.
Additional disclosure should improve the accuracy of the research record rather than create false precision.
