Responsible AI
How to Disclose AI Use in a Research Paper: A Practical Author Checklist
Document AI assistance clearly, verify its outputs and write a disclosure that follows your journal's policy.
Keep a record of the tool, task and human checks. Disclose the assistance where the receiving journal requires it; an AI tool does not take responsibility for your work.
For Indian research teams, check your university, ethics committee and receiving journal requirements alongside this workflow.
Start with the policy, not a stock declaration
A sentence saying that AI improved the manuscript is rarely enough to explain what happened. Before using a tool, read the target journal's author instructions and your institution's policy. Identify permitted tasks, restrictions on confidential material and the requested disclosure location. Save the policy URL and the date you checked it.
The ICMJE recommendations for medical journals call for disclosure at submission, including the cover letter and the appropriate part of the submitted work. They also place responsibility for accuracy, integrity and originality on human authors. This is a useful reference point, rather than a universal rule for every discipline. Read the ICMJE author guidance.
Build a small AI-use log
Our suggested working record has six fields: tool and model or version when available; date; purpose; material supplied; output used; and human verification. Keep sensitive prompts in approved storage rather than publishing them automatically. Do not upload participant information, unpublished collaborators' work or confidential review material merely to obtain a better answer.
- Language assistance: mark the passages edited and compare their meaning with the original.
- Code assistance: preserve the code version, test cases and checks on results.
- Research assistance: open every cited source and confirm that it supports the claim.
Record unsuccessful suggestions too when they matter to understanding the method. A useful log explains the workflow; it does not need to become a transcript of every interaction.
Write a disclosure readers can understand
The following is an illustrative starting point, not an official journal template. Replace each placeholder and move it to the location requested by the journal:
During preparation of this manuscript, the authors used [tool and version, if available] to [specific task] in [identified sections]. The authors reviewed and revised the output, checked [sources, code or other relevant material], and take responsibility for the final content.
If AI affected analysis or other research methods, a generic writing statement is insufficient: describe what it did, the inputs, verification and limitations needed to understand the method. Never invent a model version you cannot establish.
Check before submission
Ask a co-author to compare the disclosure against the log. Does it describe the actual assistance without understating it? Have all citations been checked against original sources? Does the final manuscript preserve the intended meaning? Has the disclosure been included in all required places?
ICMJE advises against naming AI tools as authors or treating their generated material as the primary source. Cite the underlying research instead. ICMJE recommendations.
Takeaway: transparency is a record of what you did and how you checked it. It is not a substitute for permission, good methods or accountable authorship.
Sources & editorial scope
Official sources checked on 5 October 2026. This article combines attributed guidance with ScholarVault’s suggested workflows and fictional examples. Requirements can change; check the receiving organisation’s current policy. No endorsement of ScholarVault by these organisations is implied.