When Should Researchers Disclose AI Use in Research?

At What Point Does Using AI Become Something You Must Disclose?

A researcher uses generative AI at several stages of a project. Early on, the system suggests possible research questions. Later, it helps develop analytical code, summarizes several publications, improves manuscript language, and drafts part of the discussion.

Which of these activities should be disclosed?

Should the researcher declare every interaction with AI, including minor grammar corrections? Does disclosure become necessary only when generated text appears in the manuscript? And should the researcher wait until submission before deciding what to report?

These questions have become important because scholarly publishers increasingly require transparency about AI assistance, but their policies do not always define reportable activities in the same way.

The challenge is distinguishing AI use that must be declared under a particular policy from activities that may not require formal reporting, while ensuring that consequential assistance is not concealed or mischaracterized.

Disclose AI Use When Required by Policy and When Its Role Is Material to the Research

Disclosure should be considered when the AI assistance occurs, not reconstructed from memory only after the manuscript is ready for submission. Recording consequential uses as they happen makes later reporting more accurate.

How to Determine When AI Assistance Requires Disclosure

Why Do Researchers Need to Disclose AI Use?

Disclosure allows editors, reviewers, and readers to understand how AI contributed to the preparation or conduct of research.

It may help distinguish human scholarly decisions from automated assistance and identify methodological details relevant to evaluating the findings.

For example, using AI to improve the readability of a manuscript differs from using AI to classify interview responses or generate analytical code.

The latter activities may affect how evidence was processed or interpreted, making their documentation important for methodological transparency.

Disclosure also enables publishers to assess compliance with their AI-use policies.

However, disclosure should not be mistaken for proof that the AI-assisted work is accurate. Researchers remain responsible for verifying the resulting content and methods.

Who Determines Whether AI Use Must Be Disclosed?

Disclosure obligations may arise from several sources.

Journals and publishers establish requirements for manuscript submissions. Universities may regulate AI assistance in dissertations, institutional research, or other scholarly work.

Funders and research agreements may impose additional reporting, confidentiality, or data-processing requirements.

Research ethics approvals may also constrain the use of external AI systems, particularly where participant information or confidential data are involved.

These requirements can overlap without being identical.

A researcher should therefore examine all policies applicable to the activity rather than assuming that compliance with one journal's instructions satisfies every obligation.

What Does the ICMJE Require?

The International Committee of Medical Journal Editors (ICMJE) recommends that authors disclose the use of AI-assisted technologies in submitted work and describe how they were used.

Its guidance distinguishes AI writing assistance from methodological AI use.

Writing assistance is to be reported in the acknowledgment section, while AI use for data collection, analysis, or figure generation should be described in the methods section as applicable.

The ICMJE also addresses disclosure in the cover letter and submitted manuscript.

Human authors remain responsible for the accuracy, integrity, and originality of the work.

These recommendations are influential, particularly in medical publishing, but should not be assumed to govern every journal or discipline.

What Does COPE Require?

The Committee on Publication Ethics (COPE) states that AI tools cannot be listed as authors and emphasizes transparency about their use.

Its position addresses AI assistance with manuscript writing, image or graphical production, and research activities such as data collection and analysis.

COPE calls for authors to disclose the tool used and explain its role in the materials and methods section or an equivalent section.

This approach differs in reporting location from the ICMJE's distinction between writing assistance and methodological use.

Both authorities emphasize transparency and human responsibility, but their instructions should not be treated as interchangeable.

The journal's own policy determines how the relevant guidance is implemented in a particular submission.

What Does Springer Nature Require?

Springer Nature's AI guidance distinguishes among different forms of assistance and specifies how authors should document them.

Its guidance addresses generative AI use in manuscript preparation while distinguishing defined routine AI-assisted copyediting from more substantive assistance.

Under its stated conditions, some copyediting activities do not require declaration. Other uses may require disclosure in the introduction, acknowledgments, methods, or another appropriate location.

These distinctions are important because a broad statement that every use of generative AI must be disclosed would not accurately represent all publisher policies.

Researchers should consult the current policy of the intended journal and determine whether their activity falls within an exemption or reporting requirement.

Which Types of AI Use Are More Likely to Require Disclosure?

Disclosure requirements depend on the journal and the activity, but certain uses are particularly relevant to scholarly transparency.

These are general considerations, not universal rules. The journal's policy and the actual significance of the assistance determine the required treatment.

Does AI-Assisted Grammar Correction Always Require Disclosure?

No universal rule requires disclosure of every grammar correction performed with AI.

Some publishers distinguish routine language correction from generative drafting that introduces new substantive content.

However, the boundary may be less clear than it initially appears.

A request to correct grammar can produce extensive rewriting, alter technical terminology, or change the meaning of a research claim.

Researchers should therefore examine what the tool actually did rather than relying solely on the original prompt.

Whether AI-assisted grammar correction requires disclosure depends on the applicable policy and the nature of the resulting changes.

Does AI-Assisted Translation Require Disclosure?

Translation may be treated differently from routine proofreading because it can substantially affect the wording and meaning of scholarly content.

For example, translating a manuscript from one language to another may introduce interpretive choices concerning terminology, emphasis, and disciplinary concepts.

Some publishers may require disclosure of generative AI-assisted translation, while others may address it through broader language-editing provisions.

Researchers should check the applicable policy and verify that the translated material accurately represents the original.

When translation affects research data, such as participant interviews, its methodological role may also require reporting.

The question of disclosing AI-assisted translation should therefore consider both manuscript preparation and the research procedures involved.

Does Brainstorming With AI Require Disclosure?

Not necessarily under every policy.

Researchers may use AI to explore preliminary research ideas, compare possible topics, or consider alternative manuscript structures.

Some publishers may distinguish such exploratory interactions from substantive generation of manuscript content.

However, brainstorming can become methodologically or intellectually consequential.

For example, an AI-generated suggestion may substantially shape the study design or introduce a theoretical framework that becomes central to the research.

Whether that contribution requires disclosure depends on the applicable policy and how the suggestion was incorporated.

The absence of copied AI-generated wording does not necessarily mean the assistance was irrelevant.

Does AI-Assisted Literature Searching Require Disclosure?

AI-assisted literature searching may require disclosure when it materially affects how evidence is identified, screened, or selected.

This is especially relevant to systematic reviews and other studies in which search procedures form part of the research methodology.

For example, a researcher using AI to screen abstracts should consider documenting the tool, screening procedure, human validation, and relevant limitations.

By contrast, using AI to suggest a few search terms during preliminary exploration may fall under different requirements.

The appropriate reporting should reflect the role AI played in the evidence-selection process.

Questions about AI-assisted literature-search disclosure therefore depend on methodological significance rather than the mere presence of an AI tool.

Does AI-Assisted Data Analysis Require Disclosure?

AI assistance with data analysis is particularly relevant to methodological transparency because it may influence the study's findings.

Examples include generating statistical code, suggesting analytical procedures, classifying observations, or interpreting numerical results.

Where AI materially contributes to these activities, researchers should consult the journal's reporting requirements and describe the role of the tool with sufficient detail.

Relevant information may include the analytical task, model or software used, verification procedures, and how outputs contributed to the reported results.

Disclosure does not remove the obligation to validate the analysis.

Researchers remain responsible for ensuring that AI-assisted procedures are appropriate and accurately reported.

Does AI-Assisted Qualitative Analysis Require Disclosure?

AI-assisted qualitative analysis may influence coding, theme development, interpretation, and the representation of participant experiences.

Such involvement can be methodologically consequential even when no generated text appears directly in the manuscript.

For example, a researcher may use an AI model to propose preliminary codes for interview transcripts and then revise those codes through human analysis.

Readers may need to understand how the AI suggestions were generated, evaluated, and incorporated.

Researchers should also consider confidentiality and ethical restrictions before processing participant data through external AI services.

The appropriate disclosure should accurately describe the analytical role of AI rather than implying that the tool merely assisted with writing.

When Should AI Use Be Disclosed During the Research Process?

Researchers should distinguish between documenting AI assistance when it occurs and formally disclosing it to an institution, funder, journal, or reader.

These activities may happen at different stages.

During research planning, AI use may need prior approval if it involves confidential information, restricted data, or activities governed by a research ethics protocol.

During data collection and analysis, consequential AI assistance should be documented sufficiently to support methodological transparency and later reporting.

During manuscript preparation, researchers should identify assistance that falls within the intended journal's disclosure requirements.

At submission, required declarations should be completed in the manuscript, cover letter, or submission system as instructed.

After publication, additional disclosure or correction may be necessary if an omission or error is discovered.

These stages illustrate why disclosure should be treated as part of the research workflow rather than a final administrative formality.

Should Researchers Record AI Use Even When Disclosure Is Not Required?

Maintaining a proportionate record of consequential AI assistance may be useful even when a particular journal does not require formal disclosure.

Such records can help researchers remember which tools were used, what tasks they performed, and how their outputs were evaluated.

They may also become relevant if the manuscript is submitted to a different journal with different requirements.

However, not every minor interaction needs an elaborate record unless an applicable policy requires it.

The appropriate level of documentation depends on the significance of the assistance, research risks, and institutional requirements.

The question of whether researchers should maintain records of AI use is therefore related to, but distinct from, formal disclosure obligations.

Where Should AI Use Be Disclosed in the Manuscript?

The correct location depends on the publisher's instructions and the role AI played.

Writing assistance may belong in acknowledgments or a designated AI-use statement.

Methodological assistance may need to be described in the methods section, particularly when it affects data collection, processing, analysis, or interpretation.

Some publishers require a declaration in the introduction or another designated section.

Researchers may also need to complete submission-system declarations or explain AI use in the cover letter.

The distinction between AI disclosure in methods and acknowledgments is important because a generic statement may not provide sufficient methodological information.

What Should an AI Disclosure Statement Include?

A useful disclosure should accurately describe the assistance and meet the journal's requirements.

Depending on the activity, relevant details may include the tool or model, its purpose, the research or manuscript tasks affected, and the verification undertaken by the human researchers.

Methodological AI use may require more extensive information than routine writing assistance.

Researchers should avoid vague declarations that merely state that AI was used without explaining its role when the journal requires specificity.

They should also avoid claiming verification procedures they did not perform or inventing model information they did not record.

This example should be adapted only when it accurately describes the work performed and satisfies the journal's policy.

The requirements concerning what information an AI disclosure statement should contain may differ substantially across research activities.

Does Disclosure Make Otherwise Prohibited AI Use Acceptable?

No. Disclosure and permission are separate requirements.

A journal may prohibit a particular form of AI-generated content even when its use is transparently reported.

Similarly, a research ethics approval may restrict the transfer of participant information to external services regardless of whether that transfer is later disclosed.

Researchers must therefore establish that the assistance was permitted before using it and comply with any applicable reporting requirements afterward.

Disclosure does not retrospectively authorize activities that were prohibited.

What If the Journal's AI Disclosure Policy Is Unclear?

Researchers should review the journal's current author instructions, publisher-level guidance, and submission declarations.

If the policy remains ambiguous, a concise inquiry to the editorial office may clarify whether a particular activity requires disclosure and where the statement should appear.

Researchers should retain an accurate record of the assistance and any clarification received.

They should not assume that policy silence automatically permits every form of AI use or eliminates all disclosure obligations.

At the same time, researchers should avoid inventing universal requirements that the publisher has not established.

What If AI Use Was Not Disclosed at Submission?

The appropriate response depends on whether disclosure was required, the nature of the assistance, and the stage of publication.

If a researcher discovers an omission before publication, they should review the applicable policy and contact the editorial office where correction is necessary.

After publication, an omitted declaration may require clarification or an editorial correction, depending on the circumstances.

Not every omission automatically constitutes research misconduct.

However, deliberate concealment of assistance that was expressly prohibited or required to be disclosed may raise more serious concerns.

The relevant assessment should distinguish the underlying AI use, the applicable policy, and the circumstances of the omission.

When Different AI Activities Require Different Disclosure Decisions

The example illustrates why a single statement such as "AI was used during this study" may be insufficient.

The activities differ in their potential effects on the research. Preliminary brainstorming, analytical coding, and substantive manuscript drafting may require different levels of documentation and reporting.

The correct disclosure reflects what the AI actually did and what the journal requires, rather than treating every interaction as identical.

Misconceptions About When AI Use Must Be Disclosed

How Should You Decide When to Disclose AI Assistance?

Begin by identifying what the AI tool actually did, whether the activity was permitted, and which reporting requirements apply.

Disclosure decisions should be based on the function and significance of the assistance rather than the tool's name alone.

Researchers should also consider whether an AI interaction materially affects the reproducibility or interpretation of their research.

Where it does, methodological documentation may be necessary even if the tool's output does not appear directly in the manuscript.

The appropriate standard is transparent, proportionate reporting that enables editors and readers to understand how AI contributed to the scholarship without implying that all forms of assistance have identical significance.

What to Verify Before Deciding Whether AI Use Requires Disclosure

Frequently Asked Questions About When to Disclose AI Use

Disclose AI Use According to What It Contributed, Not Merely Whether It Was Used

How to Cite This Guide

This guide is intended to be read, shared, and used in research, teaching, and academic work. If you draw on its ideas, explanations, or other content, please acknowledge the source by citing the guide. Doing so gives appropriate credit and helps your readers locate the original resource.