Generative AI was a huge mistake. We also see the flip side of this in "AI detectors" ensnaring people who are writing the way they always have. We see it in the idea that "if you use too many em dashes, then you must be AI".
I can just prompt a model to never use em dashes. Which is rather funny to me. I can make it respond like Kamina from Gurren Lagann to hype me up as well.
Btw that really doesn't work. Try with Chatgpt. It will not stop using em dashes. You will need to write code to clean it up.
Because you should instruct LLM what to do, not restrict what to not do. Tell LLM to use other symbol instead.
I use other models, but also yelling at ChatGPT typically does the job for me.