Hidden text and prompt injection finder
Text can carry characters you can't see: zero-width spaces, direction overrides, look-alike letters, and invisible "tag" characters that AI models still read. Paste text from an email, web page or document before you give it to an AI assistant.
Nothing you type leaves this page
How it works
Text can contain characters you can't see: zero-width spaces, bidirectional overrides that reorder text, letters from other alphabets that look like Latin ones, and Unicode “tag” characters. People can't see tag characters, but many AI models read them as normal text, which makes them a way to hide instructions, known as prompt injection.
The finder shows every hidden character, decodes hidden messages, lists phrases aimed at AI assistants, and gives you a cleaned copy. It keeps the joiners that Urdu, Persian and emoji need.
Questions people ask
What is prompt injection?
Instructions hidden in content that an AI assistant reads, such as a web page or email, that try to make it ignore its user and do something else, like leak data.
How do I remove zero-width characters?
Paste the text above, scan it, and use “Copy cleaned text”. It removes invisible characters and replaces look-alike letters with normal ones.
Why would text contain invisible characters?
Some are harmless, from copying out of PDFs or web pages. Others are used to watermark text, get past filters, or hide instructions for AI models.
More free tools
Found a bug or want another tool? Contact