01 · The Question
Do You Need to Tell a Research Funder That You Used Generative AI?
You use generative AI to improve the clarity of a research grant proposal, organize its objectives, or examine possible weaknesses in the methodology. Before submitting the application, you wonder whether the funding organization needs to know.
Does using AI to polish a paragraph require disclosure? What if the system helped develop the research approach or draft substantial sections of the proposal?
The distinction matters because funding organizations may have different policies concerning AI-assisted writing, intellectual contribution, originality, and transparency. Some provide specific disclosure instructions, while others establish restrictions on particular uses without prescribing a universal declaration format.
Researchers therefore need to distinguish between what a funder explicitly requires, what it encourages, and what responsible reporting may warrant when AI materially influenced the application.
03 · What You Need to Know
When and How AI Assistance Should Be Reported in Grant Applications
Why Might Research Funders Want AI Use Disclosed?
A grant application is not merely a writing exercise. It represents the applicant's proposed intellectual contribution, research commitments, and justification for receiving funding.
Generative AI may contribute in different ways. It might correct grammar without changing the underlying research, or it might suggest hypotheses, develop methodological alternatives, generate substantive arguments, or draft extensive sections.
These activities raise different questions about the applicant's contribution and responsibility.
Disclosure can help distinguish the researcher's own intellectual work from assistance provided by an AI system. It may also support compliance with funding policies and provide transparency about how consequential proposal content was developed.
However, disclosure should not be treated as a substitute for originality, accuracy, or scientific accountability.
Does Every Use of AI Require Disclosure?
Not necessarily. Requirements depend on the funding organization and the activity involved.
Some funders may distinguish routine editorial assistance from substantive AI involvement in developing the research proposal. Others may provide broader disclosure instructions or require applicants to follow institution-specific policies.
The distinction should be based on the actual governing policy, not an assumed universal rule that grammar correction is always exempt or that every AI interaction must be declared.
| Type of AI Assistance |
Possible Disclosure Consideration |
What Authors Should Verify |
| Spelling and grammar correction |
May be treated as routine editorial assistance |
Whether the funder requires disclosure of language-editing tools |
| Improving clarity and organization |
May warrant disclosure depending on the extent of assistance |
Whether AI substantially changed the proposal's presentation or argument |
| Drafting proposal sections |
Raises questions about originality and the extent of AI contribution |
Whether substantial AI development is restricted or must be reported |
| Suggesting research objectives or hypotheses |
May involve substantive intellectual assistance |
Whether the funder permits this activity and expects disclosure |
| Developing methodological alternatives |
May influence the scientific content of the application |
Whether the applicant independently justified and selected the method |
| Reviewing proposal weaknesses |
May be considered preparatory feedback |
Whether the review involved restricted information or prohibited AI processing |
| Generating references or factual claims |
Creates accuracy and research-integrity risks |
Whether all sources and claims have been independently verified |
These categories illustrate relevant considerations rather than establish mandatory disclosure thresholds. A particular funder may adopt requirements that differ from this classification.
What Does the National Science Foundation Say About AI Disclosure?
In its December 2023 notice concerning generative AI in merit review, the US National Science Foundation (NSF) distinguished between the responsibilities of reviewers and proposers.
For proposers, NSF encouraged applicants to indicate in the project description the extent to which generative AI was used and how it contributed to the proposal's development.
NSF also emphasized that proposers remain responsible for the accuracy and authenticity of their submissions, including content developed with AI assistance.
The wording is significant. An encouragement to disclose should not automatically be described as a universal mandatory declaration.
Applicants should consult the current NSF Proposal and Award Policies and Procedures Guide, the relevant solicitation, and any subsequent official guidance to determine what applies to their submission.
What Does the National Institutes of Health Say About AI-Developed Applications?
The US National Institutes of Health (NIH) has addressed AI use through policies that distinguish applicant preparation from confidential peer review.
In Notice NOT-OD-25-132, issued in July 2025, NIH established restrictions concerning research applications or sections of applications substantially developed by AI. The policy applies to applications submitted for receipt dates beginning September 25, 2025.
NIH's concern includes whether the submitted work represents the original ideas of the applicants.
This policy should not be reduced to a simple instruction to disclose AI use. If a particular form of substantial AI development is restricted, adding a disclosure statement does not necessarily make that activity permissible.
Applicants should examine the actual notice and subsequent official guidance rather than assuming that a disclosure statement resolves every concern about AI-generated content.
Disclosure and Permission Are Different Requirements
A researcher may accurately disclose an AI activity while still violating a policy that prohibits the activity itself.
Conversely, a permitted activity may not always trigger a mandatory disclosure requirement.
Permission
Whether the funding organization and other relevant authorities allow the proposed AI activity.
Disclosure
Whether and how the applicant must report the AI assistance that occurred.
Accountability
The applicant's continuing responsibility for the accuracy, originality, integrity, and feasibility of the submitted proposal.
These obligations should be evaluated separately. Researchers should first establish that their intended AI use is permissible, then determine the applicable disclosure requirements.
The broader question of restrictions imposed by research funding organizations is especially relevant when different rules govern applicants and peer reviewers.
What Should an AI Disclosure Statement Include?
When a funder requires or permits a disclosure statement, researchers should follow its specified format and location.
If the funder provides no prescribed wording, a useful disclosure may identify the AI tool, describe the relevant assistance, and explain how the applicants reviewed and retained responsibility for the final content.
The level of detail should reflect the significance of the assistance. A general statement that AI was used may be insufficient when the system materially influenced the proposal's methodology or scientific argument.
At the same time, an unnecessarily extensive account of every routine interaction may obscure the activities that genuinely matter.
Should Researchers Identify the AI Tool and Model?
Where required, researchers should report the specific tool and any additional details prescribed by the funding organization.
For consequential AI assistance, recording the provider, product, model identifier where available, and relevant dates may improve transparency and make the process easier to reconstruct.
However, this level of documentation should not be presented as a universal grant application requirement.
Researchers may retain a more detailed internal record than the funder requires them to include in the submitted application.
Where Should AI Use Be Disclosed?
The appropriate location depends on the funding program.
A funder may specify a section within the project description, provide a dedicated declaration field, or issue separate instructions about AI-assisted content.
Researchers should not automatically insert an AI statement into an arbitrary section of the application, particularly when strict formatting and page limits apply.
When the policy encourages disclosure without prescribing a location, applicants may need to determine an appropriate approach consistent with the official instructions.
If uncertainty remains, clarification from the funding organization or institutional research office may be useful.
Does AI Disclosure Remove Responsibility for Errors?
No. Disclosure identifies assistance; it does not transfer responsibility.
If an AI system invents a citation, misrepresents preliminary results, or produces an unsupported methodological claim, the applicants remain responsible for correcting the error before submission.
Similarly, disclosure does not justify presenting AI-generated ideas as independently validated scientific contributions.
Researchers should verify every consequential claim and ensure that the proposal reflects decisions they can explain and defend.
The distinction becomes particularly important when considering what must remain the applicant's own intellectual contribution.
What If the Funder Has No Explicit AI Disclosure Policy?
The absence of an explicit AI disclosure requirement should not be interpreted as permission to disregard other obligations.
Applicants must still comply with requirements concerning originality, research integrity, confidentiality, data protection, and truthful representation.
Where AI materially influenced the scientific development of the application, voluntary transparency may be appropriate, provided it is consistent with the funder's instructions.
However, researchers should avoid claiming that voluntary disclosure is universally mandatory or that it automatically resolves questions about permissible AI assistance.
Watch Out
Do not assume that declaring AI assistance makes prohibited conduct acceptable. If a funder restricts substantial AI development, confidential data processing, or another activity, disclosure does not override that restriction.
04 · A Practical Example
Deciding Whether and How to Disclose AI Assistance
Hypothetical Example
An Educational Technology Grant Application
A researcher prepares a proposal examining the effects of AI-supported formative feedback on undergraduate learning outcomes.
The researcher independently develops the research questions, selects the methodology, and establishes the proposed analysis plan. Generative AI is then used to improve paragraph clarity, organize a preliminary timeline, and identify possible weaknesses in the proposal's explanation of recruitment procedures.
Before submission, the researcher checks the funding organization's current AI policy.
Step 1: Identify the actual AI contributions
The researcher records that AI assisted with language revision, timeline organization, and preliminary critical feedback. It did not independently determine the final research objectives or methodology.
Step 2: Examine the funding policy
The researcher determines whether the funder permits these activities and whether disclosure is required, encouraged, or not specifically addressed.
Step 3: Verify the final proposal
The researcher checks that the revised text accurately represents the proposed study and that no AI-generated references, commitments, or methodological claims remain unverified.
Step 4: Prepare an appropriate declaration
If disclosure is required or voluntarily provided, the researcher uses the funder's prescribed format. Where no wording is specified, the statement describes the actual assistance without exaggerating or concealing its extent.
An illustrative disclosure, where permitted by the funder, might read:
Generative AI was used to assist with language editing, organization of the project timeline, and preliminary identification of areas requiring clarification. The applicants independently developed the research objectives and methodology, reviewed the AI-assisted revisions, and take responsibility for the accuracy and content of the submitted application.
This example is not a universally accepted disclosure template. It would be inappropriate if AI had actually developed substantial scientific content or if the funding organization prescribed different reporting requirements.
The essential principle is that the statement should accurately describe what occurred.
