AI label checker
Drop in a photo or video and see the hidden markers that trigger "AI info", "AI-generated" and "AI modified" labels: Content Credentials, IPTC source type and AI tool tags.
What this checker tells you
When you post a photo or video, the platform reads information stored inside the file before anyone sees the picture. If that information says generative AI was involved, an AI label can appear within seconds. The checker reads the same information and puts it in front of you, so you are no longer guessing.
For each file you get four things:
- A plain verdict. AI-generated marker, AI-edit marker, credentials without an AI marker, or no marker at all.
- The evidence. Every marker we found and the part of the file it sits in.
- A platform view. How Instagram, Facebook, Threads, TikTok, YouTube, Pinterest and LinkedIn are likely to treat the file.
- A privacy check. Location, serial numbers and owner name, with a one-click cleaner for photos.
The markers, explained
Content Credentials
A signed record attached by cameras, editing apps and AI generators. It lists where the file came from and what was done to it. When that record says an AI model created the content, most large platforms label it. Read more in our guide to C2PA.
IPTC digital source type
One field, a handful of standard values. trainedAlgorithmicMedia means an AI model made the whole thing. compositeWithTrainedAlgorithmicMedia means real media with an AI-made element, which is what a generative retouch writes. digitalCapture means it came from a camera.
AI tool names and generator settings
Many generators sign their work in the software or credit field. Some store the full prompt, seed and model in a PNG text block. Both are easy for a platform to read.
What the checker cannot see
Two things sit outside any metadata reader. First, invisible watermarks such as SynthID, which are woven into the pixels or audio. Second, each platform’s own visual detection, which looks at the picture itself. A clean result here means the metadata is clean. It is not a promise about what a platform will do.
Label looks wrong? If you shot the photo yourself and only retouched it, follow the fix for real photos labelled as AI. If AI made the content, see how to label it well.
Questions people ask
What does the AI label checker look for?
Content Credentials (C2PA manifests), the IPTC digital source type, AI tool names in software and creator fields, and generator settings stored in PNG text blocks. These are the markers platforms read when they decide to show an AI label.
Is my file uploaded anywhere?
No. The checker is a script that runs inside your browser. The file is opened locally and never leaves your device.
The checker found nothing but my post still has a label. Why?
Three common reasons: the platform re-encoded a different copy of the file than the one you checked, you or a collaborator switched the AI label on during upload, or the platform used its own visual detection or an invisible watermark, which cannot be read from metadata.
Does a clean result prove a photo is real?
No. Metadata can be missing for many innocent reasons, and a file with no markers can still be AI-made. The checker tells you what the file says about itself, nothing more.
Why should I check the original file and not a screenshot?
Screenshots and files forwarded through chat apps are new files with different metadata. To see what a platform saw, check the exact file you uploaded.
Does it work on phones?
Yes. It works in current versions of Chrome, Safari, Firefox and Edge on Android, iPhone and desktop.