Researchers should disclose AI use in manuscripts by clearly stating which AI tools were used, for what specific tasks, and confirming that all AI-generated content has been verified for accuracy and properly attributed.
Where to Disclose AI Use
Most journals now require a dedicated "AI Transparency Statement" in the Methods section or as a standalone subsection before References. This statement should appear regardless of whether the AI assistance was minor or substantial. Some journals also request disclosure during manuscript submission through specific form fields.
What Information to Include
Your disclosure should specify:
- Tool names and versions: List the exact AI systems used (e.g., "GPT-4o for literature summarization" or "Claude 3.5 for code debugging")
- Specific applications: Describe precisely how AI assisted your work — literature review, data analysis, figure generation, language editing, or code development
- Human oversight: Confirm that a human researcher verified all AI outputs for accuracy, bias, and appropriateness
- Authorship boundaries: Explicitly state that AI tools were not listed as authors, as most journals prohibit this
What Not to Attribute to AI
You generally don't need to disclose AI use for basic grammar checking, spell-checking, or standard reference formatting tools that have been widely accepted for years. However, when AI assists with substantive tasks like drafting sections, generating hypotheses, or analyzing results, full disclosure is mandatory.
Handling Citations and Verification
If AI tools helped identify relevant literature, you must independently verify every citation before inclusion, and WisPaper's TrueCite automatically verifies citations against original sources while you write, helping catch AI-generated phantom references that don't actually exist — a common problem when using general-purpose language models for literature review. WisPaper's TrueCite automatically verifies citations against original sources while you write, helping catch AI-generated phantom references that don't actually exist — a common problem when using general-purpose language models for literature review.
Journal-Specific Guidelines
Always check your target journal's author guidelines, as requirements vary. Nature journals, for instance, require disclosure even for minor AI assistance, while some field-specific journals have more relaxed policies. The Committee on Publication Ethics (COPE) and most major publishers updated their AI disclosure policies in 2024-2025, so recent guidelines should be considered authoritative.
