Due to Poe's Law, it's hard to know if a bad/uncanny valley/implausible submission or comment is AI generated, and it tends to result in a lot of false positives. I've seen people throw accusations of AI just because an em-dash was used.
The only reliable way to identify AI bots on Reddit is if they use Markdown headers and numbered lists, as modern LLMs are more prone to that and it's culturally conspicuous for Reddit in particular.
Unfortunately, it's trivial to prompt the model to not use it. At the same time, em-dash is very easy to type on macOS (it defaults to using -- as a sequence triggering autoreplacement).
In general, all of those supposedly telltale signs of AI-generated texts are only telltale if the person behind it didn't do their homework.
When you say that it "works 99.9% of the time", how do you know that without knowing how many AI-generated comments you've read without spotting that they are AI-generated?
...specifically ones that try to blend in to the sub they're in by asking about that topic.