How to Stop ChatGPT From Overusing the Em Dash: A Quick Guide

OpenAI says ChatGPT will now follow user instructions to avoid overusing the em dash more reliably. The change gives users better control over punctuation in outputs and addresses a stylistic pattern that has become associated with AI-generated text.

Why ChatGPT Defaulted to Em Dashes

The em dash has recently become a marker people sometimes use to identify AI-written text. Its frequent appearance across essays, emails and social posts has prompted mistaken claims and criticism. Some observers have speculated about causes, training data quirks or platform text conversions, but there’s no definitive explanation in the material shared here.

OpenAI says it has improved ChatGPT’s responsiveness to user-style instructions so the model will more reliably follow a request to limit em-dash usage. As Sam Altman wrote on X:

“Small-but-happy win: If you tell ChatGPT not to use em-dashes in your custom instructions, it finally does what it’s supposed to do!”

What OpenAI Has Changed in the New Update 

For journalists, editors and content teams, the update is a practical tool rather than a silver bullet. It lets users enforce a style preference at the model level — useful for newsroom consistency, brand voice, or avoiding punctuation patterns that trigger false AI-detection flags. But it’s not a wholesale removal of the em dash; rather, it’s a way to ask the model to reduce or avoid the character as needed.

Five simple steps to stop em-dash overuse in ChatGPT

1. Open ChatGPT.

2. On the web version, click your name → Settings → Personalisation.

3. Click the Custom Instructions text field.

4. Type a preference such as 'Please do not use em dashes.'

5. Tap Save.

The custom instructions feature is available to all users, including free-tier accounts. It’s unclear from the provided material whether the change has been rolled out everywhere in the UK and European Union.

The Limitations That Still Remain

• This setting asks the model to follow your style preference; it does not guarantee zero em dashes in every response.

• The em dash won’t vanish globally from AI outputs — you gain control over your session’s behaviour.

• Because the underlying reasons for the em-dash pattern aren’t settled here, some outputs may still contain the punctuation in edge cases or when reproducing quoted material.

For newsrooms balancing speed and consistency, the feature is a welcome control lever: enforce house style, reduce false positives in AI-detection workflows and keep copy aligned across writers. It’s a simple, user-driven fix that shifts punctuation control back toward human editors.