PT
Home / Real photo labeled as AI

Why your real photo got labeled as AI.

Phone editors and photo apps now write AI metadata into edited pictures, sometimes for edits that involve no generative AI at all. Check what your photo says about itself and remove it if the label is wrong.

Drop your photo here
JPG · PNG · WebP · MP4 · MOV · never uploaded
LosslessNever uploadedNo size limitNo sign-up

    Where the signal comes from

    Platforms label images when the file says it was made or changed with AI. Meta, for example, reads C2PA and IPTC metadata (Meta) and notes that if a photo editing tool "uses AI to change the size or color of an image, then the image may have a signal that shows it was edited using AI" (Meta Help Center). The question is which apps write that signal. Here is what the companies themselves say:

    App or deviceWhat it writes
    Google Photos (Magic Editor, Magic Eraser, Zoom Enhance)IPTC metadata indicating the photo was edited with generative AI (Google, October 2024)
    Google Photos (Best Take, Add Me)IPTC metadata indicating the image is composed from different photos, even though these features are non-generative (Google)
    Google Photos (AI edits)Content Credentials attached to any image edited with AI tools (Google, September 2025)
    Pixel 10 cameraContent Credentials on every photo created by Pixel Camera (Google)
    Samsung Galaxy S25 seriesContent Credentials for content created and edited with generative AI (Samsung, February 2025)
    Adobe Photoshop and Lightroom generative toolsC2PA metadata, which Adobe says cannot be disabled (Adobe)

    Not every Content Credentials manifest says "AI". A Pixel 10 photo carries credentials because it was captured by a camera, which is evidence for authenticity, not against it. What matters is what the manifest or the IPTC field claims, and that is what the report above shows.

    Generated, edited or composite: how the label differs

    Meta separates content "created with AI", which is labeled "AI content" on the post, from content "modified using AI", which does not get the label directly on the post, story, reel or thread (Meta Help Center). In the file this usually shows up as two IPTC values:

    • trainedAlgorithmicMedia: the whole image was generated.
    • compositeWithTrainedAlgorithmicMedia: part of a real image was changed with AI.

    If your photo shows the first value but you only removed a sign from the background, something in your editing chain recorded the edit more strongly than it deserves. That is exactly the case removing metadata is meant for.

    Fix it before you post

    1. Drop the final exported photo into the box above.
    2. Read the report: C2PA manifest, IPTC AI source flag, the tools mentioned, plus GPS and device data.
    3. Download the cleaned copy and post that. The image data is copied byte for byte, so the photo looks exactly the same.

    A label that is already on a published post won't change when you clean your file; post the cleaned copy as a new post if you need to.

    A checklist for photographers

    • Note which tools you used. A generative remove or fill leaves a different signal than a color or exposure change.
    • Check the final export, not the RAW file or an earlier version, because exporting can add or drop metadata.
    • Look at the AI source line: "AI-edited (composite)" means the file claims a partial AI edit, "AI-generated" means it claims the whole image was generated.
    • Keep a signed or untouched original archived before you share a cleaned copy.

    When cleaning won't help

    • Some AI editors also add an invisible watermark to the pixels. Google applies SynthID to images edited with Reimagine in Google Photos (Google), and Samsung overlays a visible watermark on images generated with some Galaxy AI features (Samsung). Neither is metadata.
    • If you disclosed AI use when posting, the label comes from your choice, not the file.

    Keep your original. If your camera or phone signs photos with Content Credentials, that record can later prove a picture is real. Share a cleaned copy where provenance doesn't matter, and keep the signed original as your evidence.

    Use it honestly. This is for photos that are genuinely yours. If an edit changed what a photo shows in a meaningful way, a label is fair, and many audiences expect to know.

    Frequently asked questions

    Does the Pixel 10 mark every photo?

    Google says the Pixel 10 lineup has Content Credentials built in across every photo created by Pixel Camera. That record shows the photo was captured by the camera.

    Is there a free tool to check if a photo has AI metadata?

    Yes. This page and our AI metadata checker are free, need no account and read the file in your browser.

    Why was my unedited photo labeled as AI?

    Usually some app in your chain wrote AI metadata: an AI-powered resize or color tool, a phone editor like Magic Editor, or a composite feature like Best Take. Drop the file in to see which signal it carries.

    Do non-AI features add AI metadata?

    Some do. Google says it uses IPTC metadata for images composed from different photos with non-generative features such as Best Take and Add Me.

    Is it safe to remove Content Credentials from a real photo?

    Removing them is your choice, but keep the original. A camera-signed manifest can prove a photo is authentic, so share the cleaned copy and keep the signed file.

    Will removing the metadata change the photo?

    No. Only metadata is removed. The pixels and the color profile are kept.