Northern expects faculty, graduate students, and undergraduate researchers to use AI as a tool for their scholarship, never as a substitute for their judgment or their name on the work.
Artificial intelligence can meaningfully speed up literature review, data analysis, and manuscript preparation, but scholarly work carries obligations — to research integrity, IRB compliance, and intellectual property — that AI cannot shoulder on your behalf.
This page answers common questions about using AI in research and scholarship.
Frequently Asked Questions
Can I use AI tools to help with literature reviews or summarizing sources?
Yes, but the rules depend on what the work is for. The same task carries different obligations in a course, in work headed for publication, and in reading you do for your own understanding.
In coursework. Your instructor sets the rules. Ask before using AI on an assignment and follow what the syllabus states. Permission in one course does not carry to another.
In work you may publish or submit. Journals, presses, conferences, and funding agencies set their own policies, and those policies differ. Some require disclosure, some restrict use to specific tasks, and some prohibit it. Check before you begin rather than after the draft is written.
In self-directed reading. AI tools are effective at orienting you to an unfamiliar body of work: mapping how a field frames a problem, surfacing its vocabulary, identifying where scholars disagree, and pointing you toward what to read next. Their usefulness depends on how you use them. They work well for deciding what to read, but they do not replace the reading itself.
The following practices apply in all three cases:
Choose tools that show their sources. Prefer tools that link each claim to a retrievable document or that draw only on materials that you supply. General-purpose chatbots are the wrong tool for citation work.
Verify every citation. AI tools fabricate references that look correct, with plausible authors, real journal names, and well-formed DOIs. Confirm that each source exists, then read enough to confirm it says what the summary claims. Misrepresented real sources are as common as invented ones.
Keep a record. Note the tool and version, the dates, the queries or prompts you used, what the tool returned, and how you verified it. The record is what allows you to describe your method if the work is later questioned.
In every case, you remain responsible for the accuracy of what you cite.
How do I properly cite AI use in a paper, grant, or dataset?
It is important that you provide citations when you use Artificial Intelligence (AI) in your work. There are standards set forth by APA, MLA, and Chicago for the use of AI in academic works. They are provided here from the Purdue OWL:
Artificial Intelligence (AI) How to Cite AI – Generated Content
Disclosing AI assistance with academic works is also critical. The instructions for AI usage are different from journal to journal, but there are some consistencies that should be considered when disclosing AI use in the methods section of an academic work. The CDC provides excellent and comprehensive examples on this process.
Considerations for Disclosing Generative AI Use in Scientific Work
What are the rules on using AI with unpublished data, human-subjects data, or grant materials?
Do not enter restricted, regulated, internal, institutional, or IRB-protected information into public or free AI tools. What you submit to a public tool may be retained, read by its operator, or used to train the system, and you cannot withdraw it afterward.
This includes:
- Identifiable or re-identifiable data about research participants
- Unpublished data, analyses, and manuscripts in progress
- Grant proposals before award, and material governed by sponsor terms, export control, or a Controlled Unclassified Information (CUI) designation
- Proposals and manuscripts you are reviewing for a journal, press, or funding agency
- Student records, personnel files, and other protected institutional information
Human-subjects research. Disclose planned AI use in your protocol and obtain IRB approval before you begin. If you want to add AI to a study already underway, consult the IRB first. The IRB sets the procedure and the form these requests take.
Peer review. Treat everything you receive as a reviewer as confidential. The journal, press, or agency that invited you sets the terms, and some prohibit AI use in review outright. Check their policy before applying any tool to a manuscript or proposal you are evaluating.
Sponsored work. Award and contract terms bind independently of what any tool's privacy policy states. Where a sponsor restricts disclosure, or where an export control or CUI designation applies, those restrictions govern.
A managed environment is in development. The South Dakota Board of Regents is building an institutionally-ontrolled a system-approved guarded AI environment. Keeping data inside a boundary the university governs resolves many of the privacy risks that make public tools unsuitable for sensitive material. It does not change other obligations: IRB approval, publisher and sponsor policies, and contract terms will apply to work done there as they do anywhere else. Specific capabilities, approved uses, and access procedures will be published once they are settled.
Can I use AI to help write or edit a manuscript or grant proposal?
AI tools can assist with tasks like language editing, structuring a draft, brainstorming, or summarizing your own material, but AI cannot be listed as an author. Major publishers exclude AI from authorship because it cannot take responsibility for the accuracy of the work or be held accountable — criteria only a human can meet.
As the human author of a manuscript of grant proposal, you remain fully responsible for everything in the submission, including any text, analysis, or citations an AI tool helped produce. Most journals and funders require disclosure of AI use — typically the tool name, version, and how it was used. Because these policies vary and change frequently, always check the specific journal's or funder's current author guidelines before submitting.
Plagiarism is a particular concern with AI-assisted writing.
To avoid plagiarism, the safest approach is to do your own writing and use AI sparingly, if at all, for polishing rather than generating your ideas or prose — treat it as a limited editing aid, not a substitute for your own thinking and drafting.
AI tools are also prone to fabricating citations that look legitimate but don't exist, so every reference an AI generates or suggests must be independently verified against the original source before it's included.
Additionally, because AI models tend toward generic, formulaic phrasing, relying on them across multiple manuscripts or proposals can produce oddly similar language between separate projects, which reviewers may flag as plagiarized text.
To avoid these pitfalls, treat AI-assisted drafts the same way you'd treat a co-writer's contribution: check it against a plagiarism detection tool before submission, verify every source yourself, and lean on AI for editing and structure rather than letting it generate your original content.
What ethical guidelines govern AI use in research?
Several frameworks offer reference points, though most set principles rather than binding rules:
- NIST AI Risk Management Framework — U.S. guidance for identifying and managing AI risk across four functions: govern, map, measure, manage. Increasingly used as a baseline by federal agencies and institutions.
- UNESCO Recommendation on the Ethics of AI — adopted globally in 2021; centers human rights, transparency, fairness, accountability, and human oversight.
- OECD AI Principles and, for EU-affiliated collaborations, the EU AI Act — additional international reference points.
These frameworks guide broad practice, but the rules that actually apply to your work are set locally. Check with Northern's IRB for questions about AI use in human subjects research, authorship, data integrity, or misconduct concerns — the IRB can clarify how university policy intersects with journal and funder requirements, and whether AI use needs to be disclosed in your protocol.
When in doubt, ask directly — don't assume a general framework covers your specific situation.
Who can I talk to about using AI in my specific research context?
See “Who are Northern’s AI Contacts?” under AI Definitions, Policies & Contacts