Is this image labeled as AI?
Drop in an image and see its Content Credentials, AI labels and generator tags, then remove them if you want. Nothing is uploaded.
Choose an image to check
JPEG, PNG, WebP, HEIC or AVIF. On a computer you can also drop it here.
- 1. Choose an image
- 2. Its labels are read on your device
- 3. See what it declares, remove it if you want
This checker reads what an image file declares about its own origin: Content Credentials (C2PA), the IPTC "digital source type" label, generator and software tags, and the text blocks some AI tools write into PNG files. It tells you what it finds in plain words, and can make a clean copy without any of it. The file is read in your browser and never uploaded.
What it looks for
An image can carry several kinds of provenance information, each written by a different party:
- C2PA Content Credentials. A signed manifest embedded in the file, recording which tool created the image, what was done to it (created, opened, edited, combined), and sometimes a declared source type. Generators from several large companies attach one, and so do some cameras and editing apps.
- IPTC Digital Source Type. A standard code stored in XMP metadata. The value
trainedAlgorithmicMediameans "made by generative AI";compositeWithTrainedAlgorithmicMediameans "edited or combined with AI";digitalCapturemeans "taken with a camera". - Generator and software tags. Fields such as XMP CreatorTool, EXIF Software or a credit line, which may name the tool that produced the file.
- PNG text blocks. Local image generators often save the full prompt, seed and settings in the PNG itself, under keys like
parameters,promptorworkflow.
How to read the result
| Result | What it means |
|---|---|
| Labeled as made with AI | The file declares generative AI as its source. |
| Labeled as edited with AI | The file declares that AI was used to change or combine a real image. |
| Tags point to an AI tool | No formal label, but a generator, software or text field names a known AI tool. |
| Has Content Credentials, no AI label | A provenance record exists and doesn't declare AI. Read who made it and what actions it lists. |
| Nothing found | No labels at all. This is the most common result, and it says nothing about how the image was made. |
What it doesn't do
We prefer to say plainly where the tool stops:
- No signature verification. C2PA manifests are signed. This tool shows the names in the signing certificates but doesn't check the signature or the certificate chain, so a forged manifest would look the same here.
- No pixel analysis. It doesn't guess whether an image "looks" generated. Detectors that do that make mistakes in both directions.
- No invisible watermarks. Systems such as SynthID hide a signal in the pixels themselves. That isn't metadata: this tool can't see it and re-encoding is not meant to remove it.
Removing labels and metadata
Make a clean copy redraws the image from its pixels and saves it again: JPEG stays JPEG, PNG stays PNG, other formats become JPEG. C2PA manifests, XMP, IPTC, EXIF and PNG text blocks don't survive that. The tool then scans the copy with the same checks and confirms whether anything is left.
Why remove it? Mostly for privacy. A provenance record can include an account or company name, the software you use, an edit history and thumbnails of earlier versions. A PNG from a local generator can contain your whole prompt. If you are sharing an image for what it shows, none of that needs to travel with it.
Removing a label doesn't change how an image was made. If you publish AI-generated images, many platforms ask you to say so, and it is the honest thing to do.
Checking on your own devices
You can see part of this without a tool. On iPhone, open the image in Photos and swipe up: the info panel shows the camera or app that saved it, which is sometimes enough to spot a generator. On a Mac, Preview's Tools › Show Inspector lists EXIF and IPTC fields. For GPS position, date and device, the photo metadata viewer shows everything in one place, and our guide on how to check photo metadata walks through each platform.
Where this matters
- Before publishing an image you received. A label can confirm what a source told you, or prompt a question. For newsroom practice, see journalists and sources.
- Before sharing your own generated images. Check what the file says about you and your tools.
- When cleaning a batch of photos. The EXIF remover strips metadata from many photos at once.
Photos you take for yourself don't need any of this if they never leave your phone. sigil. keeps them in an encrypted vault opened with a sign you draw, and its private camera keeps them out of the Camera Roll and iCloud Photos from the start.
Frequently asked questions
Can this tool tell me for sure whether an image was made with AI?
No. It reads what the file declares about itself. A label that says AI is a strong signal, but a missing label proves nothing: most apps and social networks strip metadata, and many generators never write it.
What is C2PA, or Content Credentials?
An open standard for a signed record of where an image came from and how it was edited. Cameras, editing apps and AI generators can attach it to the file. This tool shows what the record says but doesn't verify its signature.
What does trainedAlgorithmicMedia mean?
It is the IPTC code for an image created by generative AI. compositeWithTrainedAlgorithmicMedia means a real image was edited or combined with AI-generated parts.
How do I remove AI labels and Content Credentials from an image?
Use Make a clean copy. The image is redrawn from its pixels and saved again, which drops C2PA manifests, XMP, IPTC, EXIF and PNG text blocks. The tool then checks the copy and tells you if anything is left.
Does this remove SynthID or other invisible watermarks?
No. Watermarks hidden in the pixels are not metadata. This tool can't detect or remove them, and re-encoding is not designed to.
Is my image uploaded?
No. The file is read byte by byte in your browser. Nothing is sent to a server.