AI processing details currently live in C2PA data, which many DAMs cannot read without additional tooling. Metadata for AI exposes these details as configurable merge fields that can be embedded in delivered files and sent to integrated DAMs. This gives customers a clear, traceable view of every AI system, subfunction, and prompt that contributed to an asset.
AI Provenance Properties
Metadata Fields for AI Content are merge fields that capture AI processing details associated with an asset. The six available merge fields are grouped under AI Provenance Properties and can be used when configuring preset metadata and delivery location metadata.
These fields are compatible with both the XMP and IPTC metadata schemas, allowing AI-related information to travel with delivered assets and remain accessible in downstream systems.
Merge Field | Value |
Connector | The software agent recorded on the AI action (for example Bria, Photoroom, Gemini, Creative Force AI, Dreem) |
Subfunction | The name of the subfunction that performed the action (for example, Image to Video, Background Removal) |
AI System Used | The AI model name only, with the version removed — for a recorded model of Kling 3.5, the value is Kling |
AI System Version Used | The AI model version only, with the name removed — for a recorded model of Kling 3.5, the value is 3.5. Empty when the recorded model carries no version (for example, Bria, Photoroom, Cutout Pro, Cloudinary) |
AI Prompt Information | The prompt configured for the subfunction |
Digital Source Type | A readable label derived from the recorded action |
Preset Metadata
When configuring Image Metadata in a preset, you can find the AI Provenance Properties group in the merge-field picker. Select the fields you want to include in your metadata properties, then save the preset.
Delivery Location Metadata
The AI Provenance Properties group is available when configuring Metadata Properties for delivery locations that support Amplience, Bynder, Orange Logic, or Akeneo.
The fields are available in both single configuration mode and per-variant configuration mode. This allows AI processing details to be sent to supported DAMs along with the delivered asset metadata.
⚠️ The group does not appear for delivery methods without Metadata Properties support or in unrelated merge-field pickers, such as file and folder naming.
Practical Examples
AI metadata is written to the asset during the Asset Delivery step. When the file is delivered, the configured AI Provenance Properties appear in the file's XMP or IPTC metadata, depending on the schema selected in the configuration.
One AI Action
Asset generated by an Image to Video subfunction on Kling
Merge Field | Value Delivered |
Connector | Kling |
Subfunction | Image to Video |
AI System Used | Kling |
AI System Version Used | 3.5 |
AI Prompt Information | A model walking towards the camera |
Digital Source Type | Generated |
Two AI Actions
Background removed on Bria 1.0, and image-to-video on Kling 3.5
Merge Field | Value Delivered |
Connector | Bria, Kling |
Subfunction | Bria - Background Removal, Kling - Image to Video |
AI System Used | Bria, Kling |
AI System Version Used | Bria - Background Removal - 1.0, Kling - Image to Video - 3.5 |
AI Prompt Information | Bria - Background Removal - "a model walking, towards camera", Kling - Image to Video - "a video with model" |
Digital Source Type | Bria - Background Removal - Edited, Kling - Image to Video - Generated |
📝 Each value will be put in double quotes, separated by a comma.