In my opinion the writing style in itself becomes such an obvious tell once you've seen enough LLM output, that analysing dashes becomes pointless