I'd be shocked if the vast majority of condolence messages didn't read as “ai generated”; they're extremely formulaic. Alternatively: using an ai detector on a

I'd be shocked if the vast majority of condolence messages didn't read as “ai generated”; they're extremely formulaic. Alternatively: using an ai detector on an instance from a class of message, without doing tests on known-good and known-bad instances, is meaningless.