Semi colons - stop using them

I have GPT plus. Why does GPT ignore my ongoing request to not ever use semicolons or colons in their writing. They do it every single time even when I prompt them not to. Very frustrating.

Because it’s not a deterministic, instruction-following machine. It’s a stochastic large language model fine-tuned for a chat interface.

You need to learn to work within the confines of the system.

So just accept it will never understand my request (prompt) on semi-colons and colons?

Sure, if that’s your take on it.

The models aren’t great at handling negative prompts so it’s likely your attempts to get it to not use semicolons or colons are actually making the problem worse.

I haven’t seen an abundance of semicolons or colons myself, so I’m wondering what, exactly, the prompt is that’s causing them to appear.

Edit: I would just try something like, “use minimalist punctuation” as the instruction, don’t mention semicolons or colons at all.


this is also my experience, so perhaps in this case it could help to minimize the negativity of the request; I did a test for code and I was kind of successful with “I want you to write a semicolon-less piece of C# code willfully ignoring this part of its syntax for tic-tac-toe”,

How much text are you trying to generate? What prompt flow are you using?

Is this via API or ChatGPT? Are you using custom instructions?

I do not know about semicolon but colon always work for me, and i always get the desired result i want.

In my experience putting the words ‘Absolutely Never’ in front of things I don’t want it to do has been successful for me.

Not something you can do with the user interface, but if you are a dev, know that via the api token bias can probably solve this issue.

ChatGPT and several paragraphs at a time.

May we see your custom instructions and prompts? As others have suggested, this might be a prompting issue.

Also, you can use reflexion here after it gives you your desired several paragraphs. Once it writes something for you, ask it to “Please reflect on your answer. Do you feel that you’ve followed my instructions?”


It doesn’t have a memory, so I find myself repeating instructions and prompts over and over again.

Indeed, we are all bound by the context window for the time being.