ChatGPT explainer
What is :contentReference[oaicite] in ChatGPT text?
Those strange strings are pieces of ChatGPT's citation system that survived the copy. They mean nothing outside ChatGPT. What each one is, and how to clear them without losing your sources.
Write It teamUpdated 3 min read
If your pasted ChatGPT text contains :contentReference[oaicite:0]{index=0}, citeturn0search3 or 【4†source】, you are looking at leftovers from ChatGPT's citation system. Inside the app they turn into the small source links next to a claim. Copied somewhere else, the code can come along instead. They mean nothing outside ChatGPT, so delete them, but open the sources they pointed to first if you need to cite them.
The leftovers, one by one
| What you see | What it is | Safe to delete? |
|---|---|---|
| :contentReference[oaicite:0]{index=0} | A placeholder for a citation to a source or file the answer used | Yes |
| citeturn0search3 | A web-search citation, with invisible private-use characters around it | Yes, with the invisible characters |
| 【4†source】 | A citation to an uploaded file or a browsed page | Yes |
| ?utm_source=chatgpt.com | A tag on links that tells the site the visit came from ChatGPT | Yes, and the link still works |
Why they show up
ChatGPT writes citation markers into its answers, and the app normally turns each one into a small source link. When that does not happen, for example inside code, in some exports, or when the answer is copied through an app that takes the raw text, the marker itself lands in your text.
The citeturn markers are wrapped in private-use Unicode characters (U+E200 to U+E202) that most apps cannot display. You may see the letters with nothing around them, or empty boxes where the wrappers were, and the wrappers can stay in the text even after you delete the letters you can see.
The utm_source=chatgpt.com tag is a different thing. It is not a citation but a label for website analytics, telling the site you arrived from ChatGPT. It is harmless, but it makes links longer and shows where you found them.
Are these a ChatGPT watermark?
No. They are citation plumbing and a link label, and find and replace can delete them. They do show the text came from ChatGPT, which is why people want them gone. OpenAI's actual watermark, announced in October 2026 for ChatGPT in the EU, works through word choice and adds nothing you can see. Claude and ChatGPT now watermark their text explains how it works.
How to remove them
Check the sources first.
The markers point to sources ChatGPT used. If you plan to rely on a claim, open the source in ChatGPT and confirm it says what the sentence says. How to research with AI without fake citations walks through it.
Paste the text into a cleaner.
The ChatGPT watermark remover deletes oaicite, citeturn and 【†source】 markers with their invisible wrappers, and takes utm_source=chatgpt.com off links while leaving the rest of each link alone.
Add real citations.
Cite the sources you checked, in the style your class uses, instead of leaving a gap where the marker was.
Removing them by hand
For a few markers, find and replace works: search for contentReference and citeturn and delete each marker, and delete ?utm_source=chatgpt.com from the end of links. The citeturn markers often leave their invisible wrappers behind, and those are easier to catch with a cleaner than by eye.
Clear the citation leftovers
Paste a ChatGPT answer and copy it back without oaicite, citeturn or utm tags. Free, and nothing is uploaded.