In my case, it's often tricky because the code looks 'too perfect' at first glance, but then fails in unexpected ways.

Have you encountered this kind of AI generated code in your review? Would love to learn how others handle it.