Glossary

IPTC Digital Source Type: the field that declares AI generation

IPTC Digital Source Type is a controlled-vocabulary XMP property (Iptc4xmpExt:DigitalSourceType) whose values describe how an image was created, with trainedAlgorithmicMedia meaning generated by an AI model and compositeWithTrainedAlgorithmicMedia meaning partly AI-generated.

Code-drawn illustration for IPTC Digital Source Type

Illustration · PNG file

  1. Signature + IHDR
  2. iTXt XML:com.adobe.xmprecord
  3. tEXt parameters
  4. tEXt prompt / workflow
  5. tEXt invokeai_metadata
  6. eXIf
  7. caBX C2PA manifest
  8. IDAT pixels
  9. IEND

AI source type declared

The file labels itself as AI-made.

IPTC DigitalSourceType such as trainedAlgorithmicMedia is read from XMP and shown with its exact value.

Check an image →
AI source type declared: where this record sits inside a PNG file and what the isGenAI report says about it. The picture is a code-drawn illustration.

The vocabulary

IPTC publishes the values as URLs under cv.iptc.org/newscodes/digitalsourcetype. Camera photos use digitalCapture or computationalCapture, scans use negativeFilm or print, and AI output uses trainedAlgorithmicMedia; compositeWithTrainedAlgorithmicMedia covers mixes. Midjourney, Meta AI and Microsoft are among generators that use the vocabulary, and C2PA manifests carry the same values inside their assertions. IPTC’s 2025.1 update added further AI fields such as the AI system used and prompt information.

The field is plain XMP: editable, strippable and unsigned.

How isGenAI uses it

isGenAI parses the XMP packet, resolves the IPTC namespace properly and accepts only the exact vocabulary URLs trainedAlgorithmicMedia and compositeWithTrainedAlgorithmicMedia as declared source types. A description that merely talks about AI, or an unrelated field containing the words, is not treated as a declaration. The finding is listed as declarative evidence of moderate strength with its field named. Inside a validated C2PA manifest, the same values upgrade the record to strong direct evidence.

Why it matters for Article 50

A declared source type is one of the simplest machine-readable markings a provider can write. It satisfies the “machine-readable” part of Article 50(2) only if it is written reliably and survives the provider’s own export pipeline, which the generator pages examine vendor by vendor.

Other vocabulary values matter too. digitalCapture and computationalCapture declare camera origin, and algorithmicMedia without the trained prefix describes code-generated pictures such as fractals. isGenAI only reports the two AI values as AI declarations, so a camera value never turns into a false AI finding.

How to use this term when checking a file

  1. Upload the original file to the AI image detector or the tool named on the page.
  2. Find the finding that uses this term under “How we checked” and read the source field it names.
  3. Weigh it by kind: a signed record, a declared field, a saved prompt and a matching copy are different strengths of evidence.
  4. Remember that absence of a record is not evidence of a human author.

See the term in a real report

Frequently asked questions

What does trainedAlgorithmicMedia mean?

An IPTC value declaring that the media was generated by a trained AI model. isGenAI reports it as a declared source type.

Where is the field stored?

In the XMP packet of a JPEG, PNG, WebP, TIFF or HEIF file, under the Iptc4xmpExt namespace, and inside C2PA assertions.

Can the field be faked or removed?

Yes. It is unsigned text. That is why isGenAI reports it as moderate declarative evidence and keeps it separate from signed records.

Which generators write it?

IPTC names Midjourney, Meta and Microsoft among adopters; see the generator pages for each vendor’s documented behaviour.

Sources

Definitions follow the published standards linked above; the isGenAI behaviour described is the deployed checker as of October 2026.