ChatGPT em dashes: why they show up, and how to stop them
The ChatGPT em dash (—) is the mark people notice first. We measured it on 300 matched questions: modern AI answers used 4.5 em dashes per 1,000 words, humans used 0.21 — about 21× as many. ChatGPT at launch barely used them at all (0.06).
The ChatGPT em dash habit
This is not an old model quirk. In the original ChatGPT answers from December 2022 (same questions), we counted 0.06 em dashes per 1,000 words — almost none. By 2026 the rate had jumped, and the newest models are the heaviest users.
| Source | Em dashes / 1,000 words |
|---|---|
| gpt-6-luna (2026) | 8.06 |
| gpt-5.4-nano (2026) | 5.9 |
| gpt-4o-mini | 0.5 |
| ChatGPT (Dec 2022) | 0.06 |
| Humans (same questions) | 0.21 |
Quick remover
For the full tool — smart context, live counts, a copy button — use the dedicated remove em dashes page. A compact version is here if you want to stay on this page.
Free: 3 AI rewrites a day (up to 300 words) and unlimited quick clean. Need more? A $5 pack of 50 rewrites never expires, or go Pro at $12/month · $79/year. See plans.
Stop ChatGPT using em dashes
Never use em dashes (—) or en dashes (–). Use commas, periods, colons or parentheses instead. Do not use bold text unless I ask for it.
Paste that into ChatGPT Custom Instructions (Settings → Personalization), a Claude Project's instructions, or a Gemini Gem. For coding agents, add it to the unslop skill. Models still slip, so run the output through the remover before you publish.
Commas aren't always the right fix
- Aside: "The results — surprisingly — were flat" → "The results (surprisingly) were flat."
- Two sentences: "We shipped it — users loved it" → "We shipped it. Users loved it."
- Explanation: "One problem remains — cost" → "One problem remains: cost."
Smart context on the remover picks among those. The unslop rewrite also fixes the rest of the AI phrasing, not just the dashes.
FAQ
Why does ChatGPT use so many em dashes?
Nobody outside the labs knows for sure. Likely reasons: training on edited prose (books, journalism) where em dashes are common, and fine-tuning that rewards punchy, conversational rhythm. Our data shows it's recent: ChatGPT's December 2022 answers used 0.06 em dashes per 1,000 words; modern AI answers use 4.5.
How do I stop ChatGPT from using em dashes?
Add this to Custom Instructions (Settings → Personalization): "Never use em dashes (—) or en dashes. Use commas, periods, colons or parentheses instead." Models still slip, so paste the output into the free remover at gounslop.com/remove-em-dashes before you publish.
Is a ChatGPT em dash a sign of AI writing?
On its own, no. Plenty of writers love em dashes. But a text with several per paragraph, especially alongside bold labels and bullet lists, now reads as AI to many people.
What should I replace a ChatGPT em dash with?
Usually a comma. Use a period if both sides are full sentences, parentheses for an aside, or a colon before an explanation or list. The dedicated remover has a smart-context option that picks for you.