OpenAI starts adding hidden text tracking marks for users in Europe
The company has begun placing machine-readable identifiers inside generated text for people located in the European Union.
5 Oct 2026
OpenAI is beginning to add hidden marks to text made by its tools. These marks cannot be seen by human eyes, but machines can detect them.
The new feature is coming to ChatGPT and Codex. At first, the company is only turning on the system for users in the European Union.
The mark is called textGrain. OpenAI stated that this method matched or worked better than other ways of marking computer-generated writing.
One rival system is Google DeepMind's SynthID, which also marks text. Anthropic announced a watermarking tool in August that is based on that Google technology.
The notes end before explaining other details about the rollout. It is not clear when or if users outside the European Union will see these marks.