Assessment Design
Move from "catching AI use" to designing assessments that demonstrate authentic learning, regardless of what tools students use along the way.
This experience has 6 steps. Work through each one at your own pace. You can revisit any completed step by clicking it in the progress bar above.
Before you can redesign, you need to understand the vulnerability. Enter an assessment brief below to see how AI might attempt it, and where that approach would fall short.
Traditional assessments test whether students can produce content that sounds right. AI excels at this, including writing convincing reflections about its own output. Effective redesign shifts from policing AI use to designing for genuine engagement. The conversation itself becomes the evidence of thinking.
There's no single solution to AI-proofing assessments. Instead, consider three complementary approaches. Click each card to explore how it works and when to use it.
Show the journey, not just the destination
Instead of just submitting a final product, students submit the raw AI conversation alongside their work. The transcript is the evidence. You can see whether they challenged assumptions, asked follow-up questions, or just said "do this for me" and copied the result.
AI writes excellent reflections. A student can ask AI to critique its own output and produce a convincing "reflection" without any thinking happening. The conversation transcript is much harder to fake, and the thinking is visible in the interaction pattern itself.
Written assignments, reports, case analyses, any task where the thinking process matters as much as the output.
"Write a marketing plan" becomes "Submit your marketing plan along with your full AI conversation. Use AI as a thinking partner: ask it to challenge your assumptions, suggest factors you haven't considered, or debate whether your recommendations are realistic. I'm looking for conversation, not delegation."
Demonstrate understanding in real-time
Add live components where students must demonstrate understanding without AI assistance, not as "gotcha" tests, but as opportunities to show mastery.
High-stakes assessments, verifying understanding of complex submissions, developing professional communication skills.
"Submit a business case" becomes "Submit your business case, then present the key recommendation in a 5-minute presentation where you'll respond to questions about your analysis."
The AI Toolkit includes a full Oral Exams for an AI World guide covering course design, logistics, rubrics, and managing student anxiety. Based on Hartmann (2025).
Connect to unique contexts AI cannot know
Design tasks that require information only the student has: their experiences, their data, their specific context. AI can help, but cannot replace the personal element.
Developing professional identity, connecting theory to practice, capstone projects, applied learning.
"Analyse marketing strategy" becomes "Interview someone at an organization you're connected to about their marketing challenges, then develop recommendations using concepts from weeks 3-5."
These approaches work together. A strong AI-resilient assessment might combine personal application (Pathway C) with process documentation (Pathway A) and a brief oral component (Pathway B). Start with one, then layer as needed.
Now put it into practice. Work through this guided redesign of one of your assessments.
Strip away the format. What capability or understanding should students demonstrate?
What aspects require human judgment, personal experience, or real-time demonstration?
Select one or more approaches for your redesign:
Rewrite the assessment incorporating your chosen approach(es):
What should your rubric now emphasize? Check what matters:
Saved! Your redesign is stored in this browser.
Move from "How do I stop students using AI?" to "How do I assess what students can actually do?"
Quick wins for next semester:
"AI tools may be used as a thinking partner. Submit your full AI conversation alongside your work. I'm looking for evidence that you engaged in genuine conversation with AI, not that you delegated the task. Good conversations involve follow-up questions, challenges, and refinement."
How do you verify a research student's understanding when drafts might be AI-assisted?
Process-focused supervision:
"In our supervisions, I'll ask you to explain your work verbally. This isn't about catching AI use. It's about ensuring you're developing the research skills you'll need for your career."
Staff using AI to draft reports, emails, and documents is now common. The question is quality control, not detection.
What to look for in submissions/reports:
Ask clarifying questions about key decisions or recommendations. If someone can explain and defend their work, the tool used to draft it matters less than the quality of thinking behind it.
Our complete rubric framework for AI-integrated assessments is available in the AI Toolkit. Includes a 5-criterion template, adaptation guide, and worked examples. Filter by 'Teaching' + 'Templates' to find all four rubric documents.
๐ Want to go deeper?
Partner, Don't Police: AI in the Business Classroom, a complete guide to integrating AI into teaching with discipline-specific examples.
Shift from detection to design. Create assessments where AI assistance doesn't undermine learning
Conversation, not delegation. Design for AI as a thinking partner, and make the conversation itself the evidence of learning
Transcripts over reflections. AI writes convincing reflections. Raw conversation transcripts are much harder to fake and show the thinking as it happened
Three pathways available: Process documentation, in-class components, and personal application
Update rubrics. Value quality of conversation and reasoning over polish
Great work! Your progress has been recorded.
Checking your badge progress...
Once you've worked through the activities above, click below to record your progress.