Serious question because I keep hearing very different things from people at different schools.
At my university, the official policy says AI use must be disclosed and unauthorized AI use may be treated as academic dishonesty. But in practice I don’t think most professors are running detection on anything. The policy exists, enforcement varies wildly by department.
I know some schools have invested in detection infrastructure and are actively screening submissions. Others have the policy on paper and nothing behind it.
The gap matters because students are making decisions based on perceived enforcement, not stated policy. And if enforcement is inconsistent, the deterrence argument for having the policy at all is weaker than it looks.
What’s the actual situation at schools people have direct knowledge of? Is systematic detection happening or is it mostly spot-checks when a professor already has a suspicion?
At my institution it’s mostly suspicion-triggered. Very few faculty are running systematic detection on all submissions. The infrastructure exists but the adoption is uneven. Some departments are much more active than others. Science and engineering: almost nothing. Humanities and writing-intensive programs: more consistent use.
The policy-practice gap is a well-documented phenomenon in academic integrity research generally. Institutions adopt policies faster than they build the implementation infrastructure to support them. AI detection is a particularly sharp example of this because the tools are genuinely hard to use well and the false positive risk is high. The gap you’re observing is predictable.
From inside academia: even at schools that have paid for institutional licenses, the faculty training is often minimal. You have access to the tool but no guidance on how to interpret results, what counts as evidence, how to document a case. So even where the infrastructure is there, the practice is inconsistent.
From a people ops perspective, inconsistent policy enforcement is one of the fastest ways to undermine organizational trust. In any context, people notice when rules apply differently to different people. The fix isn’t more policy, it’s clearer implementation standards and genuine investment in training. Same problem in very different settings.
The inconsistency you’re describing creates exactly the problem you’d expect. Students who are paying attention figure out which professors are scanning and which aren’t. The deterrence only works where enforcement is visible. Uneven rollout is almost worse than no policy because it creates the appearance of fairness without the reality.