A one-line ban or a vague nod to AI both fail, because students cannot tell what to do. A workable policy is specific and tiered. Many instructors define levels: no AI use on this task, limited and disclosed use, or open use, and then label each assignment with the tier that applies. That lets a single course allow AI where it helps and exclude it where it would short-circuit the learning.

State what disclosure you expect, what counts as a violation, and where to ask if a student is unsure. Write it in plain language and put it where students will see it. A policy they understand is one they can follow.

To draft a first version quickly, try the KSARN AI course-policy generator, then adapt it to your course.