To determine whether it is a product of AI, you now have to examine more closely what you read on the Internet (and at the event in print). Sam Altman revealed on Previously, ChatGPT would ignore your instruction and continue using em dashes even if you told it not to.
People are assuming the presence of em dashes, especially if they are used in abundance, is one of the biggest telltale signs of whether something was written by a larger language model. Of course, just because a piece of text uses the em dash doesn’t mean it was actually written by an AI, but people have become suspicious of any writing that uses the punctuation mark.
This is not at all obvious while generative AI models have a tendency to distort generated text with em dashes. LLMs are trained on a large number of books and online materials, such as scientific papers, posts on public forums, and articles of the like. It is possible that punctuation appears frequently in training materials and is not flagged by AI trainers, which LLMs should avoid using.