If educators want to get to grips with students’ use of AI, they need to move from an AI detection mindset to one focused on what I call “learning verification”.
This means asking not “did this student use AI?” but “does this student understand what they submitted?”
I made the case for this shift in a previous resource entitled, “Stop investigating, start teaching”. Now I will explain what learning verification looks like in practice and tackle the question every faculty member asks first: how much time does this take?
What learning verification looks like
Verification is not a formal proceeding. It does not require you to signal suspicion or put a student on the defensive. It is an extension of what good teachers already do, which is asking students to think out loud about their work.
When a submission raises concerns, reach out directly. The message doesn’t need to be complicated: “I’d like to connect briefly to talk through your ideas on this assignment.” When you meet, ask the student to explain their thinking. What was their central argument? How did they approach the problem? Can they walk you through a specific section in their own words?
A student who genuinely engaged with the material will be able to answer these questions, not perfectly, but substantively. A student who submitted work they don’t understand will not, and that tells you everything you need to know about the grade.
- Students are asking for AI guidance, not just policy
- What your students are thinking about artificial intelligence
- Why AI literacy must come before policy
The conversation itself is often brief. Five to 10 minutes is typically enough to determine whether a student has genuine command of what they submitted. You are not interrogating; you are assessing. The framing matters: “I want to make sure I’m accurately capturing your understanding” creates an environment where an honest student feels supported rather than accused.
Document the outcome simply. A sentence in your grade book noting that a verification conversation occurred and what it established is sufficient. You are recording an educational outcome, not building a misconduct case.
The issue of time
Here is what faculty almost always say when they first hear about this approach: “I don’t have time for individual conversations with every student whose work looks suspicious.”
It is a fair concern. Early in a course, before students understand your expectations, a high number of submissions may raise questions, and adding a verification step to every one of them sounds prohibitive. But three things make the time concern better than it initially looks:
1) The detection alternative is not free. Under a detection-and-documentation approach, those same submissions each require careful review, detector outputs, written documentation, formal submission through institutional process, follow-up meetings, and usually an inconclusive result because the evidence isn’t sufficient to act on. That work rarely produces real consequences and does nothing for student learning. Verification is faster per case, and it ends in an outcome you can act on.
2) Verification has many forms. The right method should match the assignment and the nature of your concern. It might be a brief conversation, a student-submitted video explaining their approach, early drafts that show the development of ideas, a request for AI chat logs, a short written reflection, or a follow-up question through your LMS. Some forms require almost no time from you. A student submitting drafts alongside a final paper builds evidence of process into the assignment itself.
3) And the workload front-loads, then drops. Courses that implement verification see the highest volume of concerns in the first few weeks, after the first one or two graded assignments. Students are testing the environment. Once they realise that submitting work they can’t explain has direct consequences for their grade, behaviour shifts. The volume of concerns drops, often substantially, by mid-semester.
What you spend on verification is time in direct contact with your students about what they understood and what they didn’t. That time produces something: a grade you can justify, a student who understands what engagement looks like in your course, and occasionally a teaching moment that no misconduct filing ever created.
Setting the stage before the semester starts
For verification to work smoothly, set the expectation before any concerns arise. Your syllabus doesn’t need to reference AI specifically. A simple statement that you may ask students to discuss or explain their submitted work at any point in the course is sufficient. This normalises the practice, removes the element of surprise, and signals from the start that your course is about learning, not just production.
Reinforce it when introducing major assignments. Mentioning that you’re interested in students’ thinking process and may follow up with questions, describes how you teach. Students who hear it early understand what genuine engagement looks like in your course.
A practice, not a system
The premise is simple, and the practice is faster than what most faculty are doing now. The result is a clearer, more defensible read on what your students learned. You spend less time documenting concerns that never resolve, and more time in direct contact with the students themselves. The grades you assign are tied to demonstrated understanding rather than to suspicion you could not prove, and your students come to know what genuine engagement looks like in your course.
B. Jean Mandernach is the executive director of the Center for Educational Technology and Learning Advancement at Grand Canyon University.
If you would like advice and insight from academics and university staff delivered direct to your inbox each week, sign up for the Campus newsletter.
comment