University of Iowa issues guidance on using AI tools in research

University of Iowa sets framework for AI use in research

Information Technology Services (ITS) at the University of Iowa has published comprehensive guidance on the use of artificial intelligence tools in research, describing AI as both a powerful aid and a source of new responsibilities for faculty, staff, and students. The document explains how tools such as ChatGPT, Microsoft Copilot, Google Gemini, Bing AI Chat, and image generators like DALL·E can accelerate research tasks, while emphasizing that their use must remain consistent with university policies, data protection standards, and disciplinary norms.

According to ITS, AI tools can help researchers save time and resources, tackle more complex problems, and improve the quality of their work when deployed thoughtfully. The guidance frames AI as a supplementary technology that can support, but not replace, human judgment and rigorous research methods.

Two distinct modes of AI in research

The University of Iowa document distinguishes between two broad ways AI appears in research activities:

• Using AI tools for research: applying existing AI and machine learning systems as analytical or operational aids to address questions in fields such as biology, economics, astronomy, and the social sciences. In this mode, AI may assist with tasks like analysis, prediction, automation, and optimization to advance research in non-AI disciplines.

• Doing research on AI tools: developing, refining, or studying AI methods, algorithms, or systems themselves, with the primary goal of advancing artificial intelligence and machine learning as fields of inquiry.

ITS clarifies that its new guidance focuses on the first scenario—using AI to support research across disciplines—rather than on research that directly targets AI technologies.

Practical use cases in research workflows

The guidance outlines several example use cases for generative AI tools in research workflows, particularly in the early and intermediate stages of projects:

• Idea generation: AI systems can suggest related keywords, phrases, or angles that may help researchers refine topics or narrow research questions.
• Literature discovery support: AI can generate lists of potentially relevant articles, papers, and other sources that researchers can then verify and explore using conventional databases and search tools.
• Drafting titles and summaries: tools can help generate candidate titles and short summaries for papers, proposals, and other scholarly outputs, which human authors then revise.
• Drafting exploratory prose: AI can produce several paragraphs on a topic that serve as inspiration or a starting point for researchers’ own writing, although these drafts are not intended to be used as final products without substantial human review.

The University of Iowa’s view aligns with guidance from other academic institutions. For example, the University of Texas at Austin Libraries notes that chatbot-style systems such as Microsoft Copilot and Google Gemini may assist with brainstorming topics, organizing ideas, and overcoming writer’s block when used critically.[2] UT Austin’s library guidance also highlights AI’s usefulness in breaking down complex concepts or assignment prompts into more accessible language, a function that parallels Iowa’s emphasis on AI as an aid to understanding and structuring research tasks.

Transparency and academic integrity

A central theme of the University of Iowa guidance is transparency around AI use in scholarly work. ITS advises that when AI has contributed to generating ideas, text, or analysis that appears in a research output, that use should be acknowledged in an appropriate section, such as methods, acknowledgments, or another location specified by disciplinary or publisher standards.

The document stresses that AI-generated content must not be presented as wholly original work produced by the researcher alone. Instead, human researchers remain responsible for the integrity, accuracy, and originality of the final product, even when AI tools have played a role in drafting or analysis. This approach is consistent with broader academic discussions that treat AI as a tool whose contribution should be disclosed in the same spirit as other forms of assistance.

Data classification and limits on information sharing

A substantial portion of the guidance addresses data protection, tying AI use to the University of Iowa’s existing data classification framework. That framework defines several levels of data sensitivity, including:

• Public (Low Sensitivity)
• University/Internal
• Restricted
• Critical

ITS states that AI tools for which the university does not have a formal contract or agreement—such as many publicly available commercial chatbots—should only be used with Public (Low Sensitivity) data. This includes information intentionally made public, where exposure would pose minimal risk.

For AI tools that are not listed on the university’s official AI tools page, the guidance explains that no institutional agreement is in place. As a result, researchers are advised not to input University/Internal, Restricted, or Critical data into those systems. Sensitive categories that must be kept out of such tools include student records, personally identifiable information, protected health information, and research data subject to regulatory, ethical, or contractual constraints.

If researchers wish to use AI tools with higher-sensitivity data, those tools must first go through applicable university processes, including Technology and Security Review procedures. These reviews are designed to

evaluate security, privacy, contractual, and compliance risks before such tools are approved for broader use with non-public data.

Institutional support and approved tools

The guidance directs researchers to the University of Iowa’s AI tools page, which lists the AI services that the institution formally offers and specifies which data classification levels each tool is approved to handle. ITS encourages researchers who are uncertain about tool selection or data sensitivity to contact the support team at the provided email address for consultation on appropriate options.

By tying AI adoption to existing procurement, legal, and security processes, the university seeks to ensure that tools used with sensitive information meet standards for information security, privacy protection, and contractual compliance. This approach mirrors broader institutional trends in higher education, where central IT and library units are increasingly issuing frameworks and checklists for AI use.

Limitations, risks, and the role of human expertise

The University of Iowa guidance underscores that AI tools have significant limitations that researchers must actively manage. It notes that AI outputs may be inaccurate, incomplete, biased, or fabricated, and that these systems do not understand content in the same way humans do. Generative models can “hallucinate” plausible but incorrect information, making unverified outputs unsuitable as authoritative sources.

Researchers are therefore urged to scrutinize AI-generated content, verify claims against established literature or primary data, and treat AI as one input among many rather than as an arbiter of truth. The guidance emphasizes that human expertise, ethical standards, and disciplinary norms must guide how AI outputs are interpreted and whether they are incorporated into research findings.

This caution is consistent with perspectives like those from UT Austin Libraries, which highlight the need to apply critical thinking to AI-produced explanations, summaries, or suggestions and to remain aware of potential errors, biases, and gaps when relying on these systems in academic work.[2]

Intellectual property, confidentiality, and terms of service

In addition to technical and methodological considerations, the University of Iowa guidance urges researchers to consider intellectual property, confidentiality, and the terms of service of AI providers. Depending on how a service operates, text or data entered into an AI tool may be stored, logged, or used to further train the underlying model.

Because of these possibilities, ITS reinforces its recommendation against inputting sensitive or proprietary research data into tools that lack a formal agreement with the university. Instead, it advises that any AI use involving non-public data should proceed only through tools that have been evaluated and approved via institutional procurement, legal, and security channels.

Opportunity and responsibility

Throughout the guidance, the University of Iowa presents AI as both an opportunity to enhance research efficiency and a responsibility that requires adherence to institutional rules. AI is portrayed as a technology that can streamline literature discovery, support idea generation, assist with drafting, and aid analytical and predictive tasks across disciplines.

At the same time, ITS stresses that responsible AI use hinges on several practices: correctly classifying data, choosing tools that match the sensitivity of the information, remaining transparent about AI’s role in the research process, and maintaining rigorous human oversight over all scholarly outputs. Researchers who are unsure how to apply these principles are encouraged to consult with ITS or use the support channels described in the guidance.

Originally reported by Information Technology Services, University of Iowa.