How to remove the hidden AI mark from an image

What the AI mark is and when it shows up

Images created by artificial-intelligence tools come with an invisible label written inside the file: the so-called content credential, an open standard known as C2PA that records that the image was generated or edited by AI. You cannot see it by looking at the photo, but platforms can read it: it is what makes a social network show the "made with AI" notice when you post. The remove AI mark tool deletes that label from the file without redrawing the image: the pixels are copied exactly as they are, and the quality does not change.

There is also a second kind of mark, used by some generators: a signal hidden in the pixels themselves, invisible to the naked eye. This tool does not touch it. Because the image is not redrawn, a signal embedded in the picture stays there. What goes is the label written in the file's data: the one social networks read to flag the post.

How to remove the AI mark in three steps

1. Open the remove AI mark tool.
2. Upload the JPG or PNG image: one or several at once.
3. Click "Remove and download". Each image downloads as a separate file, with "-sem-marca-ia" added to the name, and the screen reports the result: "Done! We removed the hidden AI mark from the image." or "No AI mark found — the image is already free of it.".

What the tool removes (and what it doesn't)

• Removes: the content credential (C2PA) written in the file (in PNG, a dedicated block; in JPG, a segment reserved for those credentials).
• Also removes, as a bonus: location, date and device model (EXIF), editing tags (XMP), agency tags (IPTC), comments and text blocks, where some generators write the prompt used and the program name. The file comes out neutral.
• Keeps: the pixels, exactly as they were, and the color profile. Nothing is recompressed, so the look and the weight of the image barely change.
• Does not remove: a signal hidden in the pixels themselves (the invisible watermark of some generators). It would only go by redrawing the image, which this tool does not do.
• Accepts only JPG and PNG. WebP and GIF show the notice that it currently works on JPG and PNG images: convert them first with the change format tool.
• The notice "Done! We removed the hidden AI mark from the image." only appears when the content credential was there (with several images, one is enough). If the image only had other hidden data, that is deleted too, but the notice says "No AI mark found — the image is already free of it.".

When it makes sense to remove the mark

The label is not always welcome. If you edited the image heavily, mixed your own parts with what the AI generated, or simply want the photo judged by the result and not by the automatic tag, removing the mark leaves the file neutral. It also suits anyone who does not want the prompt used in generation, sometimes written in the file, traveling along with the image. Keep in mind that every platform has its own rules about AI-generated content; the tool removes the technical data, and the decision to disclose or not stays yours.

What to expect from the result

• The look of the image does not change: you are removing invisible data, not editing the photo.
• Quality is preserved, because the pixels are not recompressed. The file usually gets slightly smaller, by what was removed.
• Possible side effect: the orientation flag lives in the same data that gets deleted. A photo that only stood upright thanks to it may show up sideways: if that happens, rotate it with the rotate image tool.
• If the image went through several programs, check again after any new edit: some editors write credentials and data back when saving.

Common mistakes

• Expecting the social network to stop detecting AI in every case: if the platform analyzes the pixels, not just the label, it can still recognize the signal embedded in the picture.
• Uploading WebP or GIF: they are not accepted. Convert to JPG or PNG first.
• Assuming the image "had no mark" because the notice said so, and posting without checking: the notice refers only to the content credential. The other hidden data was deleted anyway.
• Recompressing the image afterwards in another program hoping to "clean more": that only lowers the quality; the label is already gone.

Everything happens in your browser: the image is processed on your own device and is never sent to any server.

Limit: up to 2 GB per file, with no limit on how many. Because the image is not redrawn, even large files are quick.

In short: removing the AI mark here means deleting the label written in the file's data, the credential social networks read, along with location, date and generation prompts, without touching the pixels. What is drawn into the image itself stays; what identified the file's origin does not. To strip only the hidden data from ordinary photos, documents and videos, see how to remove metadata from photos, documents and videos; and if the photo shows up sideways afterwards, how to rotate and flip an image.