E-Paper
How to get rid of em-dashes from ChatGPT results
With the rollout of OpenAI’s new GPT-5.1 model, users can finally instruct ChatGPT not to use em-dashes.
If you’re tired of ChatGPT filling your responses with em-dashes, the long dash that has practically become the bot’s calling card, there’s now a simple way to stop it. With the rollout of OpenAI’s new GPT-5.1 model, users can finally instruct ChatGPT not to use em-dashes, and this time, the chatbot actually listens.
Here’s how you can remove em-dashes from ChatGPT’s writing once and for all.
Step 1: Open ChatGPT Settings
On desktop or mobile:
-Click your profile icon in the bottom-left (desktop) or top-right (mobile).
-Select Settings.
Step 2: Go to Custom Instructions
Inside Settings:
-Tap Custom Instructions.
-You’ll see two boxes:
“What would you like ChatGPT to know about you?”
“How would you like ChatGPT to respond?”
You’ll be using the second one.
Step 3: Tell ChatGPT to stop using em-dashes
In the “How would you like ChatGPT to respond?” box, type something clear and direct, such as:
-“Do not use em-dashes in your writing.”
-“Avoid em-dashes completely. Use commas or parentheses instead.”
-“Never output the ‘—’ character.”
Save your instructions.
Sam Altman confirmed that GPT-5.1 finally respects this request, calling it a “small-but-happy win.”
Step 4: Test it
Start a new chat and ask ChatGPT to write something.
You should see:
- No em-dashes
- Cleaner punctuation
- More consistent formatting
If you still spot one occasionally, you can reinforce your instruction by reminding ChatGPT in the prompt:
“Remember: no em-dashes.”
Why this works now
Earlier models often ignored formatting instructions because language models don’t follow rigid rules, they rely on probability. GPT-5.1 improves compliance significantly, so explicit instructions now lead to better results.
Optional: Fine-tune your punctuation style
In Custom Instructions, you can also specify:
-“Use only simple punctuation.”
-“Prefer short, direct sentences.”
-“Avoid semicolons and long clauses.”
This helps produce even more human-like, clean text.
