Take-home exams and AI detection
An in-class exam answers “did they write this alone” by default. A take-home exam doesn’t, which is exactly why a detector gets reached for more often here.
Say the rule for this format specifically
A general course policy on AI assistance does not automatically answer whether a take-home exam is treated like an essay or like a closed-book test. State it on the exam itself, not just in the syllabus.
A baseline still matters, even under time pressure
Compare a take-home answer against the same student’s in-class writing or earlier timed work, not just against the class average. Writing produced quickly under a deadline naturally has different rhythm than a polished essay — that alone is not evidence of anything.
Consider what a shorter window actually reduces
A tighter deadline reduces how much revision and drafting is realistic, which can make a heavily-generated answer more likely to still read as generated rather than smoothed over. That cuts both ways: it can also make honest writing read as rushed and uneven, which is its own kind of ordinary.
If you scan, scan before grading starts
Running checks after grades are already informally decided invites the score to confirm a hunch rather than inform a reading. Build the scan into the same step as your first read, per the five-step process.