Overview
Research integrity is a core responsibility of the OU Health Campus (OUHC) research community. As artificial intelligence (AI), including generative AI tools such as large language models, become increasingly embedded in research activities, these tools must be used in ways that uphold accuracy, originality, transparency, confidentiality, and accountability. The use of AI does not alter a researcher’s responsibility to conduct research ethically, appropriately credit sources, protect confidential and proprietary information, and comply with institutional and sponsor requirements.
In the OU AI Research Guidance for Researchers, artificial intelligence is defined as a machine-based system that can generate outputs with varying levels of autonomy and may be used through institutionally supported systems, enterprise software, or public-facing platforms. While responsible AI use may enhance efficiency and innovation, OUHC researchers remain fully responsible for all AI-assisted content. Inappropriate, undisclosed, or careless use of AI may compromise the research record and, in some circumstances, raise concerns related to research misconduct.
Research Misconduct, FFP, and AI
The OUHC Ethics in Research Policy defines research misconduct as fabrication, falsification, or plagiarism (FFP) in proposing, performing, reviewing research, or reporting research results.
- Fabrication is making up data or results and recording or reporting them.
- Falsification is manipulating research materials, equipment, processes, or data such that the research is not accurately represented.
- Plagiarism is the appropriation of another person’s ideas, processes, results, or words without giving appropriate credit.
Importantly, misconduct can occur whether committed directly by a researcher or through the use of tools, assistants, or technologies, including artificial intelligence systems. Use of AI does not shield a researcher from responsibility for FFP. Examples of AI-related risks include:
- Submitting AI-generated text or analyses that misrepresent originality
- Using AI to fabricate or alter data, images, or results
- Incorporating AI-generated text without appropriate verification or attribution
- Relying on AI outputs that introduce false, misleading, or non-citable content
Allegations of research misconduct are assessed under OUHC and federal procedures, regardless of whether AI was involved.
Federal Sponsor Guidance on AI Use
Federal research sponsors, including the National Institutes of Health (NIH) and the National Science Foundation (NSF), have issued agency-specific requirements governing AI use in proposal development, peer review, and research activities.
National Institutes of Health (NIH)
NIH has issued several notices clarifying expectations related to AI use in grant applications and peer review:
- Originality of Applications (NOT-OD-25-132): NIH does not consider applications—or sections of applications—that are substantially developed by AI to be original. Identified AI-generated content may be referred for further review or enforcement action. Investigators should not rely on AI to generate scientific ideas, aims, or substantive proposal narrative.
- Prohibition in Peer Review (NOT-OD-23-149): NIH strictly prohibits peer reviewers from using generative AI tools to analyze grant applications or draft critiques. Uploading applications, critiques, summary statements, or review discussions into AI tools constitutes a breach of peer review confidentiality.
- NIH Peer Review FAQs: NIH has reaffirmed that AI use in peer review poses confidentiality and security risks and may result in corrective action if policies are violated. OUHC faculty and staff serving as NIH reviewers must not use AI tools at any stage of NIH peer review activities.
National Science Foundation (NSF)
The National Science Foundation issued a Notice to the Research Community on Artificial Intelligence and revised its research misconduct policies to explicitly address AI.
- AI Included in the Definition of Research Misconduct: Effective December 8, 2025, NSF revised its definition of research misconduct to include fabrication, falsification, or plagiarism committed directly or through the use of AI-based tools in proposing or performing research, reviewing proposals, or reporting results.
- Use of AI in Proposals: NSF permits limited use of AI in proposal preparation, but investigators remain fully responsible for the accuracy, originality, and integrity of all content. Researchers should carefully consider whether use of AI should be disclosed and should avoid sharing proposal materials with public or third‑party AI tools that may retain or reuse data.
- Merit Review Restrictions: NSF strictly prohibits reviewers from uploading proposals or review materials into non-approved AI systems. This prohibition is intended to protect confidentiality and the integrity of the merit review process.
NSF’s overall approach is not to ban AI, but to ensure its use aligns with ethical standards, transparency, confidentiality, and long-standing norms of scientific integrity.
Tools, Detection, and Researcher Support
iThenticate Access Through the RIO Office
OUHSC researchers may obtain free access to iThenticate through the Research Integrity Office (RIO). iThenticate can be used to screen manuscripts and proposals for textual similarity and potential flags associated with plagiarism or AI‑generated content.
While no tool can definitively determine AI authorship, similarity and pattern flags can help researchers identify areas requiring closer human review before submission. Access and instructions are available here:
Researchers are encouraged to:
- Review sponsor and journal AI policies before submission
- Use detection tools as a screening aid, not a determination
- Retain drafts and documentation showing responsible authorship and review
Best Practices for Responsible AI Use in Research
The following best practices are aligned with the University of Oklahoma AI Research Guidance for Researchers and the OUHC Ethics in Research Policy.
- Accountability and Responsibility: Researchers remain fully accountable for all content, analyses, code, images, and conclusions produced with AI assistance. AI outputs must be carefully reviewed, verified, and validated before being incorporated into research records, proposals, manuscripts, or presentations.
- Limited, Purpose‑Driven Use: Use AI only when it clearly adds value to the research activity and where its capabilities and limitations are well understood. AI should not be used simply because it is novel or convenient, and should never replace scholarly judgment, critical analysis, or independent verification.
- Documentation and Disclosure: All use of AI in research should be documented in research records (e.g., lab notebooks, methods sections, SOPs, or acknowledgements). When disseminating research, clearly disclose:
- the name and version of the AI tool used,
- the purpose and stage(s) of use (e.g., analysis, drafting, visualization), and
- how human oversight and verification were applied.
Researchers also must also comply with sponsor‑, journal‑, and conference‑specific disclosure requirements.
- Originality and Attribution: Ensure that research outputs remain original and appropriately attributed. AI tools must not be used to generate plagiarized text, images, or ideas, or to obscure authorship. Follow applicable citation standards for acknowledging AI use, consistent with disciplinary and publisher requirements.
- Accuracy, Bias, and Limitations: Critically evaluate AI outputs for errors, hallucinations, and bias. Researchers should understand known limitations of AI tools, assess potential bias (particularly in data involving human participants), and mitigate bias through human review, validation, and transparent reporting.
- Data Privacy and Confidentiality: Do not input sensitive, confidential, or restricted information into unsecured or public AI systems. This includes protected health information (HIPAA), student records (FERPA), human subject data, proprietary or unpublished research, export‑controlled information, and data subject to special consent or governance requirements.
- Appropriate Image and Data Handling: AI must not be used to fabricate, falsify, or inappropriately manipulate research data or images. Any AI‑assisted data processing or image modification must preserve the accuracy and integrity of the research record.
- Use of Approved Systems and Oversight: Whenever possible, use AI systems that are institutionally supported or approved. Use of non‑approved AI tools with OU data may require prior IT security assessment and institutional approval.
- Training and Awareness: Researchers are encouraged to participate in available AI ethics, integrity, and best‑practice training and to remain current on evolving institutional, sponsor, and publisher expectations.
References and Resources
Disclaimer: Microsoft 365 Copilot was used to assist the authors with organization and clarity. This document was created to summarize central issues relevant to the use of AI in research and research integrity and is not intended to be comprehensive. Researchers are encouraged to directly consult other primary sources of information such as the references and resources linked above.