Scan Images for AI-Generator Metadata

Most AI generators leave fingerprints in the file — if nobody re-encoded it. Thirdwatch's Image Provenance Checker checks media files for the strongest public authenticity signals — C2PA Content Credentials manifests, EXIF software tags, and embedded AI-generator markers — and returns a labelled verdict per file.

Skip the setup: Run this as a ready-to-go task on Apify — pre-loaded with the configuration from this guide.

Why scan for AI markers

'Is this image AI-generated' is a moderation, research, and marketplace-integrity question with no reliable eyeball answer. But generators write their names into the file — software tags, parameter chunks, workflow metadata.

aiSignals lists the exact markers matched, so the verdict is evidence, not a score.

The checker parses the file's metadata layers directly, so a URL in means a provenance verdict out — with the evidence attached.

How does this compare to the alternatives?

Why this Actor

• Real C2PA parsing — manifest count, labels, generator, signers, actions, assertions.
• 30+ AI-generator signatures across EXIF, PNG text chunks, and embedded metadata.
• Full EXIF extraction — camera make/model, timestamps, software tags.
• aiSignals lists exactly which markers matched — the verdict is auditable.
• Honest semantics — absence of signals is 'not detected', never 'proven human'.

How to do it in 3 steps

Step 1: Configure the input

Set the inputs as shown below — urls takes the targets, direct media URLs — JPEG, PNG, WebP, GIF and MP4 all work.

Step 2: Run the Actor

Run it from the console, the API, or the linked saved task. One dataset row is written per media item.

Step 3: Use the output

Each row carries hasC2PA, the parsed manifest details, EXIF fields, embedded text chunks, the matched aiSignals, and the aiLikely verdict.

{
  "urls": [
    "https://upload.wikimedia.org/wikipedia/commons/3/3f/JPEG_example_flower.jpg"
  ]
}

Each dataset row looks like:

{
  "url": "https://example.com/image.jpg",
  "format": "jpeg",
  "sizeBytes": 132518,
  "hasC2PA": true,
  "c2pa": {
    "manifestCount": 7,
    "labels": [
      "c2pa.claim",
      "c2pa.signature"
    ],
    "generators": [
      "Adobe Firefly"
    ]
  },
  "exif": {
    "Software": "Adobe Firefly"
  },
  "aiSignals": [
    "exif_software:Adobe Firefly"
  ],
  "aiLikely": true
}

What to watch for

A clean result means no detectable signals — metadata is easily stripped, so treat aiLikely: false as 'nothing found' rather than 'authentic'. C2PA manifests are parsed and reported, not cryptographically verified.

Related use cases

• Check an Image's C2PA Content Credentials
• Audit a Photo's EXIF for Authenticity Signals
• Verify Listing Images Before Trusting Them
• Run a Batch Provenance Report Across an Image Set
• All Thirdwatch use-case guides

Run the Image Provenance Checker on Apify Store — pay per result, free to try, no credit card to test.