Charlie HillsFree guide · Prompts & Workflows
The ChatGPT Watermark: What It Means for Your Work
What ChatGPT's EU text watermark means for your work, what it can't tell anyone, and 5 prompts that sort your side of it in about 30 minutes.
Charlie Hills Added 3 November 2026 · Free
What’s inside
- 1What actually changed
- 2What a watermark can and can't tell anyone
- 3Six questions you will get
- 4Your 30 minutes
- 5Make it permanent
- 6When something goes wrong
- 7Sources
What actually changed
|
Where |
What is happening |
Status |
|
ChatGPT and Codex in the EU |
OpenAI is adding an invisible watermark to eligible text on all plans. It said on 05/10/2026 this would roll out over the coming weeks. |
On |
|
ChatGPT outside the EU |
OpenAI says it is not making text watermarking a global default at launch. |
Off for now |
|
The OpenAI API |
It stays off by default. You can turn it on for select models, in one project or across your organisation. |
Your choice |
|
Short answers and code |
The EU code of practice does not require a watermark under 200 tokens (about 150 words) or in code snippets. |
Not required |
|
Images and audio |
A separate system. Anyone can check a supported file at openai.com/verify. |
Already live |
That’s the first part. The rest of this guide is free too.
