Combining the various "tells" of an LLM (em dashes, delve, grammatical signs etc) with the context (Reddit comments vs professional setting), you could establish a rough probability it was AI generated. At this point, it's the best we can hope for.