If you copy and paste from ChatGPT, is there a watermark?

No. Pasting ChatGPT's answer into an essay or an email doesn't carry a watermark across with it, because there is no ChatGPT text watermark to carry. What can come across is stray Unicode, which is a real thing and a different thing.

The watermark part

OpenAI built a text watermarking method and never deployed it. As of September 2026, text from ChatGPT — free, Plus, or through the API — carries no statistical mark, so there is nothing for a "ChatGPT watermark detector" to find.

That may not hold forever. OpenAI has signed the EU's Code of Practice on Transparency of AI-Generated Content, which asks signatories to mark AI-generated text, and has said it wants to extend provenance signals to its text models. It has not said when, or how, and nothing is running today.

The characters part

Separately, ChatGPT output sometimes contains characters you can't see. The common one is a narrow no-break space, codepoint U+202F, sitting where an ordinary space should be. People reported it in o3 output in 2025 and again in GPT-5, mostly in long answers. OpenAI's position is that it's an artefact of training rather than a deliberate mark.

It does survive copy-paste, which is the part that affects you:

• a search box won't match text you just pasted
• a CSV column splits in the wrong place, or a compiler rejects a line that looks fine
• someone running a checker will see it, and may read more into it than it supports

Finding one proves nothing about who wrote the text. Word, InDesign, PDF exports and ordinary French typography all produce the same character, and plenty of ChatGPT output has none in it at all.

How to check what you pasted

Paste the text into the hidden character checker. It lists every invisible codepoint it finds, tells you what each one is, and gives you the text back clean if you want it gone.

If what you actually want is the longer answer about marking, it's here: does ChatGPT watermark text?