What's going on with take-home assessments?
Over the last couple of years, AI tools have gone from being inconsistent in their ability to complete homework tasks, but in 2026, that's no longer the case. Unproctored quizzes, essays & presentations are easy work for these tools, and the data shows that many students are using them to complete work outside of class.
So, what do we do about take-home assessment? In August we ran a webinar to explore that question. A lot of folks have been asking us for the details, so we decided to make our recommendations & the full recording available to everyone.
If you're still figuring out what to do about your take-home assessments this Fall, read on!
what we recommend
Changes you can make today
A Berkeley study this year looked at 500,000 course grades and found something important (though probably unsurprising). Grade inflation from AI tracked, almost perfectly, with how heavily a course weighted homework. Courses that weighted homework at 40 percent inflated far more than courses that weighted it at 10 percent.
That was the starting point of our webinar, and it sets up the checklist we walked faculty through. Five changes, about three hours end to end, all of them things you can do before your students arrive.
1. Re-weight before you redesign. (30 min.) The single highest-leverage change on the list, and the one most instructors skip because it feels too small to matter. You don't need new assignments; you need new point values. One table in your syllabus can shift enough weight off take-homes and onto proctored or supervised work to close most of your exposure by itself. Do this first, it has the biggest impact.
2. Write an AI clause per assessment, not per course. (45 min.) One blanket AI policy at the top of a syllabus is a policy no one reads, and no one can enforce. What works instead is three short sentences, one attached to each assignment, that put the student in one of three lanes: AI use is restricted, AI is allowed (with disclosure) or AI is required. The lane changes from assignment to assignment, and that is exactly why students remember it.
3. Convert one assignment so that it requires use of AI. (60 min.) Pick one take-home this term where the critique of an AI output is what earns the grade, not the output itself. The strongest version is the planted-error audit. You feed the model a prompt, plant a specific mistake in the response, and the student's job is to find it, name it, and correct it. It is a genuinely hard exercise, and it is only possible in a world where every student has an AI tool at hand.
4. Add one process artifact to your biggest take-home. (15 min.) The take-home that carries the most points. Ask for the working file, or a two-question build note describing how the student got there and what they'd change on a second pass. It costs almost nothing to add and is unexpectedly diagnostic. A well-written build note tells you within a paragraph whether the person in front of you understands what they turned in.
5. Plan anything you can supervise now. (10 min.) A short in-person viva, or a 30-second model check at the front of a scheduled meeting, is enough to anchor the rest of the term. This is the quiet failure on the list. Rooms and time slots run out by week three, so the item gets pushed to week four, and then week eight, and then the semester is over. Get the calendar invites on the books this week.
The framing we keep coming back to is that every assessment you give this fall can be one of three things: AI-visible, AI-irrelevant, or AI-integrated. All three are achievable before the semester begins. What is not achievable is AI-proof. That word has to be removed from the vocabulary.
If you only have an hour this week, do items 1 and 4. Together, they close most of the exposure and buy you room to do the rest once the term begins.
Worth a look
Watch the full webinar recording here
If the topic of this email is relevant to you, check out the full 45-minute webinar here.
Skills are beginning to matter more than credentials for employers
A trend that's raising important questions about the role of universities and the shape of curricula.
What on earth does Madagascar have to do with the Industrial Revolution?
Always proofread your work, folks!
Sam