How can a forum label machine-created posts in structured data?

The Citead team, updated

In short. A forum can use digitalSourceType with an IPTC URI to describe the digital source of a DiscussionForumPosting.

Definition: digitalSourceType is a Schema.org property that records an IPTC digital-source code for a CreativeWork. A forum can place that property on a DiscussionForumPosting object to describe whether the post's digital content came from a trained model, a pure algorithm, an algorithmic alteration, or a composite process.

Key facts

  • Google describes DiscussionForumPosting as markup for user-generated forum posts, and says not to use it for content primarily authored by a website's publishers or their agents. Google's Discussion Forum structured data documentation states this in its content guidelines.
  • Schema.org defines digitalSourceType as an IPTC code indicating the nature of the digital source for a CreativeWork. The Schema.org property definition expects an IPTCDigitalSourceEnumeration value.
  • Schema.org describes DiscussionForumPosting as a posting to a discussion forum and places it under CreativeWork in the type hierarchy. The Schema.org type definition lists CreativeWork as its parent type.
  • The IPTC vocabulary defines trainedAlgorithmicMedia for digital media created algorithmically with an Artificial Intelligence model trained on captured content. IPTC's Digital Source Type vocabulary provides the code and definition.
  • The same vocabulary defines algorithmicMedia for media created purely by an algorithm without sampled training data, and algorithmicallyEnhanced for algorithmic modification or correction that leaves the main content unchanged.

How does the property fit a forum post?

Google describes discussion forum markup as intended for forum-style sites where people collectively share first-hand perspectives. It defines DiscussionForumPosting as an original posting that is the topic of discussion, while a posting can consist of text or media content. The guide also says that each post should include its entire text when that text is present.

Schema.org describes DiscussionForumPosting as a CreativeWork subtype, while its digitalSourceType property describes the nature of the digital source for a CreativeWork. Taken together, those definitions provide the vocabulary relationship for recording source provenance on a forum-post object. The property describes how the digital content was made; it does not replace the post's author field.

Google's forum documentation separately requires author, datePublished, and either text, image, or video for the post type. Treat digitalSourceType as additional source metadata alongside those fields.

Which value separates trained-model content from other algorithmic content?

IPTC's vocabulary distinguishes the algorithmic cases by how the content was produced:

ValueUse described by the vocabulary
trainedAlgorithmicMediaThe media was created algorithmically using an Artificial Intelligence model trained on captured content.
algorithmicMediaThe media was created purely by an algorithm without sampled training data, such as software using a mathematical formula.
algorithmicallyEnhancedAn algorithm modified or corrected the media without changing its main content, with the process initiated or configured by a human.
compositeWithTrainedAlgorithmicMediaThe media was augmented, corrected, or enhanced using a trained model, including inpainting or outpainting operations.

The vocabulary also includes composite for a mix of several elements. Schema.org notes that IPTC codes are not declared mutually exclusive in general, although some combinations can be contradictory or considered incompatible by the maintainers of the definitions.

How should the markup be written?

Use the IPTC concept URI as the value of digitalSourceType. This fragment shows the trained-model value inside a forum-post object:

{
  "@context": "https://schema.org",
  "@type": "DiscussionForumPosting",
  "digitalSourceType": "http://cv.iptc.org/newscodes/digitalsourcetype/trainedAlgorithmicMedia"
}

The fragment does not contain the other fields required for the complete Google forum markup. Add it to the same object that describes the post, then preserve the complete post text and the author and publication information required by Google's guide. If the source is pure algorithmic media, replace the URI ending with algorithmicMedia; for a human-configured algorithmic alteration, use algorithmicallyEnhanced.

FAQ

Is digitalSourceType a required Google forum field?

Google's forum guide lists author, datePublished, and post content such as text, image, or video in its type definitions. digitalSourceType adds source information from the Schema.org and IPTC vocabularies.

Which value identifies content made with a trained model?

Use trainedAlgorithmicMedia when the digital media was created algorithmically with an Artificial Intelligence model trained on captured content. The IPTC vocabulary publishes that definition with the code.

Which value describes an algorithmic edit to existing content?

Use algorithmicallyEnhanced when an algorithm modifies or corrects the media without changing its main content and a human initiated or configured the process. Use compositeWithTrainedAlgorithmicMedia when the enhancement uses a trained model.

Can more than one IPTC source code apply?

Schema.org says the codes are not declared mutually exclusive in general. It also notes that some combinations can be contradictory or considered incompatible by the maintainers of the definitions.

Sources

  1. Google Search Central, "Discussion forum (DiscussionForumPosting) structured data", 8 September 2026
  2. Schema.org, "digitalSourceType", undated
  3. Schema.org, "IPTCDigitalSourceEnumeration", undated
  4. Schema.org, "DiscussionForumPosting", undated
  5. IPTC, "Digital Source Type", 2024

Check how a page reads to answer engines. Open the GEO score analyzer

Test a page and see what to fix first

The analysis usually takes less than 10 seconds and needs no account.

400,000+ pages scored in the public benchmark