After an incident that ChatGPT falsely included hidden characters into some of their texts there is the
possibility that users everywhere use these texts without knowing they have invisible watermarks planted
inside. To prevent such an occurrence from happening this tool can identify invisible Unicode characters
and delete them for you.
Especially for students that work with ChatGPT, Gemini and Depseek for their studies as well, it might be
a good idea to double check before a thesis or something like that is somehow watermarked digitally!!
Use the tool with caution tho. There is not a 100% guarantee that this is reliable and meets the hidden
watermark that AI uses. Maybe AI is even better detectable by choice of words, so this only covers Unicode
watermarks.
Text is represented as a sequence of Unicode characters. A wide variety of characters exists, including some
that remain invisible in standard text. Researchers have demonstrated that invisible Unicode characters can be
embedded in AI-generated responses as a form of watermarking, since they typically do not occur in genuine
human-authored content.
By detecting and removing these characters to the best of my ability, I try to ensure that your
texts remain authentic and uncompromised.
The application is implemented in PHP, and no personal data—including any submitted text—is stored permanently.
All submissions remain strictly confidential.
For transparency, the site initiates a session and temporarily retains the submitted text solely to prevent the browser’s
“Resubmit form data” warning when reloading a page containing a previously submitted form, ensuring a seamless user
experience.
For anyone interested in reviewing the tool’s source code, it is publicly available on GitHub.
Even if I respect your privacy I still advice you to generally never upload any sensitive data to any webpage on the internet. If you need to check sensitive data host this page locally on your computer. You can find instructions for this on Github.