AI metadata checker: see what your file tells platforms
Find out whether a photo or video carries C2PA Content Credentials, an IPTC "AI-generated" flag, generator names or GPS location before you post it. Checked in your browser, never uploaded.
What the checker looks for
Every file gets a four-line verdict at the top of its report, followed by the full list of metadata found. The four lines are the signals that matter most for AI labels and for privacy:
| Verdict line | What it means | Where it is found |
|---|---|---|
| C2PA Content Credentials | A signed provenance manifest is attached to the file | JPEG APP11 segments, PNG caBX chunk, WebP C2PA chunk, MP4/MOV uuid box |
| AI source flag | An IPTC digital source type says the content is AI-generated or AI-edited, or a PNG holds generation settings | Inside the C2PA manifest, in XMP, or in PNG text chunks |
| GPS location | Coordinates of where the photo or video was taken | EXIF, or location boxes in MP4 and MOV files |
| App or generator names | Well-known tool and company names mentioned in the metadata | C2PA manifests, XMP and video tags |
Below the verdict, the report lists every metadata block the parser found, with its type and a short description, so you can see exactly what would be removed.
How to read the result
"Likely to trigger an AI label"
You see this when the file carries a C2PA manifest, an AI source flag, or both. These are the signals platforms say they read. Meta labels images on Instagram, Facebook and Threads by reading the "AI generated" information in the C2PA and IPTC standards (Meta), and TikTok reads C2PA Content Credentials to label AI-generated content uploaded from other platforms (TikTok).
The AI source flag tells you which kind of claim the file makes. The IPTC value trainedAlgorithmicMedia means the whole image was generated by a model. compositeWithTrainedAlgorithmicMedia means AI was used to edit part of an otherwise real image, which is the situation behind many complaints from photographers whose edited photos were labeled.
"No file-level AI signal found"
The file has no C2PA manifest and no AI source flag. That is useful to know, but it is not proof that the content is not AI-generated. A watermark woven into the pixels, such as Google's SynthID, is not metadata and does not show up in any metadata checker (how SynthID works). Platforms can also apply labels when content is disclosed by the person posting it.
Why names are listed as "mentions"
When the checker sees a name like Adobe, OpenAI or Google inside a manifest, it reports that the name is mentioned. It does not verify the cryptographic signature, so it cannot tell you who signed the file. A name can appear because that company made the tool, because an earlier version of the file was edited with it, or simply as text in a field.
Check before you post
The most common reason to use a checker is to avoid surprises after publishing. A quick routine:
- Drop the exact file you plan to upload, after all your edits and exports. Metadata often changes when you export, so checking an earlier version tells you little.
- Read the verdict. If it says the file is likely to trigger an AI label, look at the AI source flag line to see whether it is a full "AI-generated" claim or an "AI-edited" one.
- Decide. If the image is genuinely AI-generated, many platforms expect you to disclose it, and keeping the label is the honest choice. If it is your own photo that was only retouched, or you simply don't want to share GPS and device details, use the button under the report to remove everything and download a clean copy.
- Check the cleaned copy again if you like. Dropping it back in should show "None" on every line.
Other ways to check a file
| Tool | Good for | Trade-off |
|---|---|---|
| This checker | A quick answer on AI signals and GPS, for photos and videos, in batches | Does not validate C2PA signatures |
| Content Credentials Verify (contentcredentials.org) | Reading the full manifest and validating the signature | You upload the file to the site |
| c2patool | Printing the complete manifest store as JSON on the command line | Needs installing, C2PA only |
| ExifTool | Listing every EXIF, XMP and IPTC tag, for example with exiftool -a -G1 photo.jpg | Command line; the output is long and technical |
Formats and limits
- Supported: JPG, PNG, WebP, MP4, MOV and M4V. AVIF, HEIC and GIF are in progress.
- Large videos: only the small index of a video is read, so multi-gigabyte files are checked in moments.
- Timed metadata tracks: some phone videos store per-frame data, such as live location, as a separate track. The checker reports it, but it cannot be removed without re-muxing the file.
- No signature validation yet: the checker detects manifests and reads their AI flag and named tools, but does not check them against trust lists.
Want to clean files rather than inspect them? The AI label remover does both in one step, and the C2PA page explains what is inside a manifest.
Frequently asked questions
Is the file uploaded to check it?
No. The file is read by your browser and analysed with JavaScript on your device. Nothing is sent to a server.
Can this tell whether an image was made with AI?
It tells you what the file says about itself: whether it carries C2PA Content Credentials or an IPTC AI source flag. It does not analyse the picture, and it cannot detect pixel watermarks such as SynthID.
My real photo shows an AI-edited flag. Why?
Editing apps can record that a generative or AI-assisted tool was used on part of the image, using the IPTC value compositeWithTrainedAlgorithmicMedia. The checker shows that value on the AI source flag line.
Does it check videos too?
Yes. MP4 and MOV files are checked for C2PA and XMP boxes, location and device tags, and the names of the apps that exported them.
Is it free?
Yes. No account, no daily limit and no file size cap.