There's a movement to reclaim it, and LLMs will probably use it less and less as time goes on

I understand your frustration. It helps me personally to remember that it's not quite GenAI that is the problem—it's the greed that surrounds its implementation/usage. GenAI could have been just a tool, occasionally useful, but our obsession with productivity and the capitalistic goal of replacing workers to get even more profit for the owning class—or in other words just personal greed—is the real problem, not GenAI.

So are words. 🤷‍♀️ Em dashes are only well-represented in LLM output because they’re well-represented in the training data. There’s no simple litmus test, the scummy LLM companies would just tweak their models if there was one.