From now on, ChatGPT will be harder to hide. It turns out that with the release of the new o3 and o4-mini models, OpenAI chatbot started leaving hidden watermarks — Unicode characters in the generated texts in place of regular spaces.
You can see hidden characters using online tools or text editors such as Google Docs or Sublime Text.
If you enter the text generated by the chatbot in Word and press ctrl+shift+8, normal spaces will turn into dots (.), while ChatGPT watermarks will appear as circles (°).
Interestingly, OpenAI says that this is a bug, not a feature — but does not specify whether they plan to remove it. Older models do not contain such watermarks.
How to remove watermarks?
It’s actually very simple: open any text editor that displays special characters and replace them with their standard counterparts — as shown in the Rumi team’s video guide.
Interestingly, the idea of deliberately adding watermarks to ChatGPT-generated products has been discussed for a long time, although the company itself is in no hurry to update it (we can assume that it will not have the best effect on the chatbot’s popularity). Earlier, OpenAI opened free access for students, who can obviously be one of the most active users of the chatbot.
It is worth reminding that the models o3and o4-mini, released by OpenAI last week, are the first AI models that «think with pictures». Earlier, the company releasedspecial AI models for IT professionals GPT-4.1, GPT-4.1 mini, and GPT-4.1 nano, and in the coming weeks we expect to release o3-pro — a version of o3 that uses more computing resources to generate answers exclusively for of ChatGPT Pro subscribers.
Don’t thank ChatGPT — OpenAI spends «tens of millions of dollars» on your courtesy
Spelling error report
The following text will be sent to our editors: