Skip to content
AI Watermark Remove

Check if your AI text is detectable

AI tools hide characters in the text they write. You cannot see them. The checker your university runs can. Paste your text below and see exactly what is in there. Then remove it if you want.

0 characters
More options
Checking is free: 5 texts and 2 images a day, no signup.

Checking your text

    Three steps, about ten seconds

    No queue, no account, no upload bar to watch.

    1. 01

      Paste your text

      Or drop in an image. We read it, answer, then throw it away. Nothing is saved and nothing is trained on.

    2. 02

      See what is hidden

      Not a percentage. A list: what each character is, how many there are, where it sits in your text. This part is free.

    3. 03

      Take a clean copy

      One click. The words are identical. Only the invisible parts are gone. This is the part a pass pays for.

    What we look for

    Not guesses about your writing. Specific characters: named, counted, shown in place.

    U+200B
    Zero-width space
    Takes up no space and survives copy and paste.
    U+2060
    Word joiner
    Sits between words and shows nothing at all.
    U+E0000
    Tag characters
    A whole hidden message can be written in these.
    U+FE00
    Variation selectors
    On their own, a known way to hide bytes in text.
    U+00A0
    Odd spaces
    Spaces that are not really spaces, where a normal one belongs.
    C2PA
    Content Credentials
    A signed record inside an image naming the tool that made it.

    Two problems, one engine

    Documents leak through characters nobody can see. Images leak through records written inside the file. Pick the one you actually have.

    What we cannot do

    Most tools here will not tell you this, so we will. There are two kinds of AI watermark and only one of them can be removed.

    Hidden characters come out completely. Zero-width characters, tag characters, private-use characters, C2PA, EXIF, XMP. These are just data sitting in your file. We find all of it and delete it, and you can check again to prove nothing is left.

    Word-choice watermarks do not. Google's SynthID works by nudging which words the model picks. The watermark is the word choice. There is no character to delete and no public tool can read it, because that needs Google's key. Anyone selling you SynthID removal for text is selling you nothing.

    Images split the same way. We take the hidden data out completely, and we tell you plainly when a picture also looks like it carries a mark in the pixels that we cannot touch.

    Work out which problem you have before you pay anyone to fix it. Checking here is free, so that is easy.

    Read the detail

    Common questions

    Is checking really free?
    Yes. You see the whole report: every hidden character, what it is called, how many there are, where it sits. Five texts and two images a day, with no account. You only pay if you want the clean copy.
    Why is the removal the paid part?
    Because nobody should have to pay to find out whether they have a problem. If we charged for the check, you would be paying for a number you cannot verify. This way you see the evidence first, and people whose text turns out to be clean pay nothing.
    Does ChatGPT put invisible characters in text?
    OpenAI has never said it does it on purpose. But ChatGPT text very often arrives with invisible characters in it anyway: odd spaces, zero-width characters, narrow no-break spaces picked up while formatting. On purpose or not, they are in your document. They survive copy and paste, and anyone looking will find them. We show you which ones are there instead of guessing why.
    Does Gemini watermark its text?
    Yes, with SynthID, but not with hidden characters. SynthID for text works by nudging which words the model picks, so the watermark is the word order itself. No character-removal tool can find or remove that, including this one. Reading it needs Google's key. Gemini text can carry hidden characters on top of that, and those we do remove. Be very sceptical of anyone selling SynthID removal for text.
    Will this get me past my university's AI detector?
    Probably not, and we would rather say so than sell you a promise. Turnitin and tools like it mostly look at writing style, not hidden characters, so taking the characters out usually will not change their answer. What we do is tell you exactly what is in your file. That is real, checkable information about your own document. It is not a detector bypass, and anything advertised as one is overselling.
    Is it safe to paste my essay here?
    Your text is read in memory and thrown away when we answer. We do not write it to disk, we do not log it, and we do not train anything on it. The only thing we keep is a salted hash used to count your free checks, and it cannot be turned back into an IP address. If you would rather not send it at all, the whole method is written out on the How it works page.
    What is C2PA, and why is it in my image?
    C2PA, also called Content Credentials, is a signed record inside an image file saying what made it. Most big image generators now add one automatically. It survives uploads and downloads, and usually re-saving too. Anyone can read it with a free verifier. We take it out along with EXIF and XMP, without re-encoding your picture, so nothing about the image itself changes.
    Can you remove SynthID from an image?
    No, and nor can anything else that leaves the picture intact. SynthID for images lives in the pixels rather than the metadata, and it survives cropping, resizing, compression, even a screenshot. Removing it would mean regenerating the image, which changes what it looks like. We remove the metadata completely and tell you plainly when a picture looks like it also carries a pixel watermark we cannot touch.
    Do I need an account?
    No, for either tier. Checking needs nothing at all. If you buy a pass, that browser is unlocked the moment you pay. No password, no key to type in. The confirmation page also gives you a link that unlocks your phone or another computer, if you want it there too.
    All questions

    Check your document now

    It takes ten seconds and costs nothing. If it comes back clean, you have lost ten seconds.