Post #2665042
2026-05-12 13:54 UTC
@avuko@infosec.exchange @christopherkunz@chaos.social @pikesley@mastodon.me.uk Em dashes are just one of many red flags one can use to determine if something was actually written by a human. If someone is throwing out the whole thing because someone used a single em dash, that person is being stupid. In fairness to the person you replied to, they also mentioned more than one potential sign in this blog post.
Me personally, if I see writing that uses em dashes like a child who discovered a new word and is using it everywhere, it won't be enough for me to hand-wave it as generated, but it makes me suspicious. If the post is also long-winded without actually saying anything (and I mean *anything*) of substance and also aggressively uses bullet points, my experience has been it almost certainly wasn't written by a human.
Replies (1)
-
@avuko@infosec.exchange 2026-05-12 14:39
@disorderlyf@todon.eu @christopherkunz@chaos.social @pikesley@mastodon.me.uk Yeah, and to add insult to injury: bullet lists, but every bullet is an emoji (with at least one being “🚀”). 🤨 My only real use of LLMs currently is to rephrase something I’ve written so neurotypical people can understand or accept it without too much resistance. My main problem with “AI detection”, perhaps more in the beginning than currently (as the slop is becoming very recognisably sloppy), was that a disproportionate amount of ND people found their writing to be flagged as suspicious, especially by teachers etc. Often because of their careful use of typographical symbols, punctuation marks, and grammar.