Custom GPTs within ChatGPT Edu offer faculty and staff new ways to customize AI tools for particular tasks, projects, or groups. While the ability to create a custom GPT is already available, we have now enabled the ability to share custom GPTs externally, making it possible to share these tools beyond the university environment. With this added capability, it is important to understand appropriate data use, sharing settings, and access considerations before making a Custom GPT available to others.

What Is a Custom GPT?

A Custom GPT is a personalized version of ChatGPT Edu that you design to help with a specific task or set of questions. You can think of it as creating a focused AI assistant that follows instructions you provide and uses only the information you choose to include. The term GPT (generative pretrained transformer) refers to the type of technology behind ChatGPT that allows it to generate responses based on patterns in language, but you do not need to understand the technical details to use it effectively.

If you are familiar with creating an agent in Microsoft Copilot, a Custom GPT serves a similar purpose. In both cases, you are defining how the tool should behave, what it should help with, and what information it should draw from. Rather than starting from a blank chat each time, a Custom GPT gives you a reusable assistant designed for a particular need.

While Custom GPTs can be helpful, they are still AI tools. Users remain responsible for reviewing, validating, and applying judgment to all outputs, especially before sharing information, making decisions, or using results in university communications or processes.

Within the ChatGPT Edu environment, Custom GPTs can support practical university use cases such as answering common questions about internal processes, summarizing policy or guidance documents, helping draft routine communications, supporting onboarding or training activities, or guiding users through multi-step tasks using approved, non-sensitive information.

Data Classification: Public and University Internal Only

All content used in or shared through a custom GPT must follow University of Iowa data classification expectations.

Custom GPTs should only include or be pointed towards data classified as University Public or University Internal. Examples include publicly available university information, published resources, or internal process documentation intended for faculty or staff to use. For more information on the criteria that make information public or internal, visit the University of Iowa Data classification guidelines and Institutional Data Policy web pages.

Custom GPTs must not include restricted or sensitive data, such as personal data, protected health information, student records, or other data classified above University Internal, unless explicit approval has been granted through appropriate university processes. These requirements apply regardless of whether a GPT is shared internally or externally. Users are responsible for ensuring that confidential data is not shared inappropriately outside of the university. Check the AI tools page for information on the data classification level of other AI tools.

How and Why to Create a Custom GPT with ChatGPT Edu

Log in to ChatGPT Edu 
Faculty and staff who have a license to ChatGPT Edu should access it using their HawkID@uiowa.edu. This ensures work takes place within the university’s enterprise environment and aligns with institutional expectations for data use. If you do not already have a license, you can request one by talking to your supervisor or IT leadership. For more information, visit the ChatGPT Edu service page.

Create a Custom GPT 
When using the Custom GPT builder in ChatGPT Edu, you can define:

• The purpose and behavior of the GPT
• Instructions or guardrails for how it should respond
• Optional reference materials or context (limited to Public or University Internal data)

Custom GPTs are useful when you want more consistent responses, reusable workflows, or a focused assistant designed for a specific audience or task. 

Sharing a Custom GPT: Choosing the Right Settings

Before sharing a Custom GPT, test it thoroughly to ensure responses are accurate, appropriate, and aligned with how you intend the tool to be used.  When sharing a Custom GPT, it is important to carefully select the appropriate sharing option.

Internal sharing 

Sharing within the university workspace ensures that access is limited to authenticated university users and is generally the safest option for tools intended for internal audiences. You can choose to share with specific people that you identify, or anyone at the university who has the link.

External sharing (“Anyone with the link”) 

Selecting “anyone with the link” allows access by anyone who obtains the link, including individuals outside the university. When this option is used, you lose control over who can access the GPT and how it may be used or shared further.

Because of this, Custom GPTs shared externally should be reviewed carefully to ensure they contain only University Public or University Internal data and do not expose internal processes or information beyond their intended scope. Custom GPTs should only be shared with others that also have permission to access to the data contained within them.

Always choose the most restrictive sharing option that meets your needs.

Support and Training

Consultation on creating and sharing Custom GPTs, understanding data classification in relation to building custom GPTs, and selecting appropriate sharing settings is available. For questions or guidance, contact the AI Support Team at ITS-AISupportTeam@uiowa.edu. Want to get more updates about ChatGPT Edu at Iowa? Subscribe to the AI at Iowa newsletter.