The IRB has seen an increase in the use of AI to generate study materials. All AI-generated content must be carefully reviewed by the research team before submission. Investigators remain responsible for ensuring that submitted materials accurately reflect the study and comply with UW-Madison requirements (this includes student projects).  

When reviewing AI-generated content, pay particular attention to the following: 

  1. Accuracy, consistency, and clarity. Lack of consistency is a recurring issue and delays study approval. Focus on consistency, while clearly describing what your study will do, rather than including statements about what it will not do. 
  1. Specificity. Avoid vague descriptions such as “IRB-approved procedures will be followed.” Applications should clearly explain who will perform study activities, what they will do, and how participants will move through the study. 
  1. Institutional requirements. Confirm that recruitment, consent, or study procedures align with UW-Madison policies and guidance. 
  1. Practical feasibility. Ensure study timelines and processes are feasible and realistic in practice. 
  1. Consent form language. Use UW-Madison consent templates when possible and ensure all required institutional language is included. 
  1. Responses to reviewer comments. When responding to requested changes, concise responses are often most helpful, such as confirming edits were made while making changes in the body of the application.  

Before entering study information into an AI tool, also consider the following: 

  • Industry sponsors may prohibit the input of confidential information into AI tools. Be familiar with any contractual restrictions that apply to your study. 
  • Ensure that the AI tools you use are appropriate for your data classification and use university approved tools. Consider whether entering information into the tool could expose confidential information, sensitive data, or intellectual property. 

If you choose to use AI, a careful review of its generated content is imperative. Submissions that are unclear, inconsistent, or that include logical impossibilities delay study approval.