By skyforbes I think everyone's aware by now how prevalent the em dash "—" is with AI writing, and for some god forsaken reason, no matter how much you tell it not to use it, it will use it pretty much immediately.
GPT 5 for example using it immediately in the 3rd sentence: https://chatgpt.com/c/6916ba9d-56f8-8326-90fc-b0e9cbb6e4d0
However GPT 5.1's prompt adherence / instruction following is SO GOOD that it actually… listens to you for once!
GPT 5.1: https://chatgpt.com/c/6916baed-5de0-832b-a0ee-c45897becb69
And yes you don't need to put it in the prompt, you can put it in custom instructions (it was just easier to share the chats this way)
I'v seen a lot of people misunderstand what GPT 5.1 is. It doesn't matter if you want AI to be more robotic, or if you don't like GPT 5.1's tone and you like GPT 5 or 4o or 4.1 better. With how much of an improvement there is to instruction following, you can just TELL it what to do and what not to do. Don't like lists? Tell it not to use them.
It actually LISTENS now – heck it doesn't even use em dashes if you tell it not to! Do you guys recall GPT 5's non stop asking followup questions at the end of each response "if you'd like I can turn this report into a PDF with charts, graphs and a cherry on top", that no matter what you couldn't get rid of even with multiple instances of memories or custom instructions telling it not to? Well now 5.1 listens.
You can truly customize your ChatGPT however you like. THAT is the point of the 5.1 update.
On a sidenote, 5.1 is perhaps the most censored ChatGPT model they've made so far so…