When using Jetbrains AI assistant I've found that when the "Generate Documentation" produces something that doesn't look correct at first it is always something broken in my code.
When the way AI describes what some code does doesn't jive with what I intended the code to do, the bug has been in the code I wrote every time so far.
Obviously this process isn't perfect. Sometimes the documentation it produces is too vague or too detailed. Sometimes it adds unimportant details and omits important ones. But when it produces something that is completely different than what I expected, then it is a huge red flag.