Last updated: August 9, 2026
Generative AI Assessments
There are both academic and practical reasons you may choose to incorporate generative AI assignments into your course. For example, you may believe that AI will be a skill needed in the students’ future careers in your field. Perhaps you see AI as a tool to help students deepen their understanding of and engagement with your content. Or you may develop AI assignments to discuss its ethical and academic implications, instead of as a learning tool. One side effect of exploring how to integrate AI into assignments is that instructors begin to think more deeply about how to design assignments that cannot be easily generated by AI alone to deter plagiarism and cheating.
Note, it is critical to develop AI policies for your course along with policies for specific AI assignments.
AI Resilient Assignments
If you are interested in creating assignments that AI cannot complete easily, some basic approaches include:
- Breaking down longer assignments with feedback opportunities to support revision
- Developing real world, personalized, and reflective assignments that have more than one solution
- Emphasizing teamwork, in class assignments, and alternatives to writing assignments such as oral exams.
Refer to these guides by the University of Chicago , UCLA’s HumTech group and Columbia University for guidance in developing AI-resilient assignments.
Considerations for Developing an AI-Enabled Assignment
Alignment with Your Course Goals
In the development of AI assignments, the primary consideration is whether the use of AI will help your students achieve the learning goals of the course. Ask yourself, does this assignment help student gain skills and knowledge central to your course and field? Furthermore, consider whether the assignment is engaging enough to warrant incorporating AI. Are you asking students to go above and beyond the AI-generated content? An impactful assignment will challenge students to transform, expand upon, correct, or critique the information and content generated by AI or learning about themselves in relationship to AI. Educational pedagogy expert Derek Bruff gives further insight into how to think about AI assignments as they relate to course design in his blog post about AI and writing assignments.
Guidelines for Use
If you integrate AI into your assignments, be sure to discuss your expectations with your students. It is essential that they understand why you have decided to allow AI in the course and its role in their learning. Furthermore, students can be engaged in wider conversations about AI and its personal impact on their lives. The University of Calgary has developed a set of recommendations of how to start these conversations. One strategy is writing a code of conduct that emphasizes critical thinking and sets guardrails of proper use. You can provide a prewritten list of guidelines or work with the students from scratch by posing questions about AI and learning.
For example, the class may have guidelines such as:
- We will only use AI to help our intellectual development, not replace it.
- We will be transparent in our use of AI.
- We will not submit AI generated text without attribution.
- We will follow guidelines of when AI is appropriate to use.
Assignment Mechanics
Detailed instructions for an AI assignment will raise the chances for a successful learning experience. Students are not familiar with the processes of this novel type of intellectual work, and thinking through the different facets of the activity will help you to execute and evaluate the assignment confidently.
First, consider adopting a tiered scale for assignments in your course. This framework can be adopted and a specific tier can be shared with students to quickly communicate the level of AI allowed for each assignment. Another layer for developing a successful assignment is creating an outline for each assignment that has instructions for the assignment, as well as AI guidance such as prompt examples, allowed tools, and citation expectations. Finally, be prepared to offer feedback on their disclosed AI use to help guide them as they build their skills.
AI Literacy
Both you and your students should have a level playing field when it comes to understanding generative AI. You cannot count on students to understand the pitfalls and limitations of AI or even how to use the tools. There are existing resources on AI literacy developed specifically for students that can be a starting point. This library guide from the University of Arizona instructs students on AI, plus there is a companion guide for instructors as well.
Not all students may take to generative AI equally and will not have the skills to architect effective prompts for your discipline or type of assignment. You can support them by modeling prompt generation and working with AI in class before assigning AI for outside assignments.
Ethical Concerns
There are ethical issues to consider when using AI for assignments beyond questions of plagiarism and academic integrity:
- To minimize threats to the privacy of your students and yourself, be sure you are using a Duke supported generative AI tool
- Do not upload personal or sensitive information into the AI model. Although the models that Duke supports do not train public AI models, identifiable information should remain confidential.
- Students who do not have sufficient funds for subscriptions to AI tools can be at a disadvantage. Let students know that Duke provides no cost access to ChatGPT and Microsoft Copilot to all university users.
Note, for instructors who do allow AI in assignments, there should be considerations for students who do not want to use it on ethical grounds. This can be solved by making AI engagement low-stakes or optional.
Generative AI-Enabled Assignments
Below are some general considerations for incorporating AI into your courses. For more specific guidance, we encourage you to seek out examples from your discipline or related to the core skills of your course. For inspiration, some resources worth exploring are the resources from the AI Pedagogy Project (sortable examples of assignments by discipline and skill) and Stanford’s AIMES Library of Examples.
Tutoring may be its best use
Generative AI can be a coach for learning when used correctly. Students can easily get more information about what they don’t understand and additional practice on skills and concepts they haven’t mastered. By the same token, it can assist instructors who need to provide additional assistance to students and are pressed for time to find resources. Instructors can get ideas for teaching a skill or subject with activity descriptions and lesson plans. AI can generate practice problems or review questions for exam prep, which frees up time for instructors for other class prep.
There are also positive gains in equity when generative AI is used in a tutoring setting. A neurodiverse student may find conversations with a bot to be non-judgmental and less stressful when needing help. Non-native speakers can ask for word and concept definitions to level up their understanding of course content and context. The review and tutoring capabilities of AI can help all students to practice concepts and receive feedback on their progress.
When instructors provide guidance and sample prompts it can help students understand the line between getting study help from AI versus letting AI complete their assignments. You can also point students to Duke Academic Resource Center’s AI toolkit which teaches students about using AI responsibly as a study aid.
Writing assignments are the most problematic
Incorporating AI into writing assignments is not without peril. Students’ reliance on AI content could potentially lead to losing skills in academic writing. There is the risk that students might mistakenly believe that AI is inherently better at developing ideas and expressing information; leaving students uncomfortable adding their own voice to writing. Without training on how to check the validity of AI content and conduct independent research, students may miss out on how to evaluate sources and compare ideas.
The line between writing support, editing, outlining, brainstorming, drafting and cheating is hard to discern and the borders differ across disciplines and contexts. As educators, it would be short sighted to not account for these tensions as we (re)design writing assignments. Amy Stornaiuolo from University of Pennsylvania suggests in her framework for approaching AI assignments that AI can assist students with feedback on their writing, act as a partner to spark creativity, and shed light on writing rhetoric. On the other hand, AI interferes with the development of students’ voices and its content and ethical implications must be considered through a critical lens.
Looking for specific ideas for updating writing assignments in the age of AI? The University of Iowa’s Writing Center has a list of possible AI-enabled writing assignments. Derek Bruff, from the Center for Teaching Excellence at University of Virginia, describes the questions he asks himself when redesigning his writing assignments. Finally, writing instructor Anna Mills’ open source resource can help you discuss the pitfalls and potentials of AI with your students.