Find and remove hidden characters in AI text
AI assistants can leave characters you cannot see in copied text: narrow no-break spaces, zero-width joiners, soft hyphens, unusual dashes and quotes, hidden citation codes. Often called AI watermarks, they make text easy to flag and break search, code and forms. This tool lists every one by name and removes them.
How to find and remove hidden characters in AI text
- Paste the AI output. The scan runs instantly in your browser. Not even the text leaves the page.
- Read the character report. Every unusual character is listed with its Unicode name, code point, how many times it appears and what it does.
- Remove and copy. Hidden characters are deleted, special spaces become normal spaces, lookalike letters become real ones, and quotes and dashes are normalised if you want.
Why use myowntext for this
Nothing is uploaded
Text is processed by your browser. There is no server, so contracts, cover letters and private notes never travel anywhere.
Instant, even offline
Every result appears as you type. Once a page has loaded it keeps working without a connection.
Shows its work
Counts, highlights and reports tell you exactly what was found and what changed. Nothing is hidden behind a spinner.
No account, no limits
Free with no sign-up, no word cap, no ads and no nag to upgrade.
Questions people ask
What is an AI text watermark?
Hidden or unusual characters that end up in text copied from AI chat apps: narrow no-break spaces (U+202F), non-breaking spaces, zero-width characters, soft hyphens, curly quotes and em dashes, plus the private-use characters that wrap citation codes in web-search answers. They look like normal text but are easy for software to spot, and they break search, code, forms and spreadsheets. This tool shows each one by name and removes it.
Which characters does it look for?
All Unicode format characters and joiners, every space variant other than a normal space, soft hyphens, byte order marks, bidirectional marks, variation selectors, private-use characters, unusual dashes and quotation marks, Cyrillic and Greek letters standing in for Latin ones inside words, citation markers, and any other non-printing code points. Each is listed by name.
What are lookalike letters?
A Cyrillic а or о is drawn exactly like the Latin letter but is a different character, so a word containing one fails search and spell check and is a classic marker in text. The report lists each one, and the fix swaps it for the real Latin letter. Genuine Russian or Greek words are left alone.
Does it change my words?
No. Only the characters between and around the words are replaced or removed. Sentences stay exactly as written.
Is the report accurate for code?
Yes, and it is a good reason to use the tool: a zero-width space inside a variable name or a narrow no-break space in a command will break code silently.