The IPTC Photo Metadata Working Group is proposing a draft set of properties for recording details of images created using generative AI systems. The group presents a draft of these fields for your comments and feedback. After comments are reviewed the group intends to add new properties to a new version of the IPTC Photo Metadata Standard which would be released later in 2025.

Use Cases

The proposals detailed here are intended to address these scenarios, among others:

• How do you know which system/model generated this image? For instance, if you wanted to compare how different systems—or versions of systems—interpret a given prompt, where would you look?
• How can you know what prompt text was entered, or image shared as a starting point? If you want to recreate similar images in the future with the same look, where should that info be stored?
• How can you tell who was involved in the creation of a generative AI image? 

Example Scenario

You are the new designer for an organisation, and need to create an image for a monthly column. You are told to use generative AI, but your boss wants the end result to have the same “look and feel” as images used previously in the column. If you needed to find those that were published previously—what information would be most useful in locating and retrieving the image(s) in your organization’s image collection?

Proposed Properties:

• AI Model
  Name: AI Model
  Definition: The foundational model name and version used to generate this image.
  User Note: For example “DALL-E 2”, “Google Genesis 1.5 Pro”
  Basic Specs: Data type: Text / Cardinality: 0..1

• AI Text Prompt Description
  Name: AI Text Prompt Description
  Definition: The information that was given to the generative AI service as “prompt(s)” in order to generate this image.
  User Note: This may include negative [excludes] and positive [includes] statements in the prompt.
  Basic Specs: Data type: Text / Cardinality: 0..1

• AI Prompt Writer Name
  Name: AI Prompt Writer Name
  Definition: Name of the person who wrote the prompt used for generating this image.
  Basic Specs: Data type: Text / Cardinality: 0..1

• Reference Image(s)
  Name: Reference Image
  Definition: Image(s) used as a starting point to be refined by the generative AI system (sometimes referred to as “base image”).
  Basic Specs: Data type: URI / Cardinality: 0..unbounded

All of these properties are of course optional, not required, but we would recommend that AI engines fill in the properties whenever possible.

Request for Comment

The intent is for a new standard version including these fields to be proposed at the IPTC Autumn Meeting 2025 in October to be voted on by IPTC member organisations. If approved by members, the new version would be published in November 2025.

Please send your comments or suggestions for improvements using the IPTC Contact Us form or via a post to the public iptc-photometadata@groups.io discussion list by Friday 29th August 2025.