Disclosing AI-assisted content

01Is there schema for AI-generated content?

There is no dedicated schema for AI-generated content as such, and no Google rich result that rewards declaring it. What exists, as of November 2026, is narrower and more useful: one schema.org property that describes the digital source of a creative work, an embedded IPTC metadata field for images, and Google guidance on when a disclosure helps readers.

An AI disclosure is a statement, visible to readers, that explains how AI was used to make a piece of content. Structured data can repeat that statement in a form machines read, but it cannot replace it. Google's structured data guidelines already require markup to describe what is visible on the page, so a disclosure that exists only in JSON-LD is the wrong way round.

The rest of this guide covers what Google has said, what schema.org can express, how bylines should work, how images are labelled and a pattern that keeps all of it consistent.

02What does Google say about AI content?

Google's position is that quality matters more than production method. Its February 2023 post, Google Search's guidance about AI-generated content, says appropriate use of AI or automation is not against its guidelines, while using automation to generate content primarily to manipulate search rankings breaks its spam policies. The same post answers three questions that shape any disclosure plan.

• Bylines: consider accurate author bylines wherever readers would reasonably ask who wrote the content.
• Disclosures: AI or automation disclosures are useful where someone might ask how the content was created; add them when they would reasonably be expected.
• AI as author: giving AI an author byline is probably not the best way to make clear to readers that AI was part of the process.

Google's newer documentation page on using generative AI content on your website goes further on accuracy. It says AI output may contain inaccuracies and that all AI-generated content should be fact-checked before publishing, and it extends that review to metadata such as titles, meta descriptions, structured data and image alt text. It warns that using generative tools to produce many pages without adding value for users may violate Google's spam policy on scaled content abuse, and it suggests giving readers context about how automation was used, including image metadata. For ecommerce, it notes that Google Merchant Center requires AI-generated images to carry the IPTC DigitalSourceType TrainedAlgorithmicMedia value.

So the practical reading is simple. Disclosure is a reader courtesy that Google endorses, not a ranking factor that Google rewards, and structured data written by AI needs the same human check as the article it describes.

03What can schema.org express today?

Schema.org has one property built for this job. digitalSourceType is defined on CreativeWork and takes a value from the IPTCDigitalSourceEnumeration, a schema.org mirror of the IPTC Digital Source Type vocabulary. Because Article, BlogPosting and ImageObject are all subtypes of CreativeWork, the property is valid on each of them.

Two limits are worth stating plainly. Schema.org's own usage panel for the property shows fewer than 1,000 domains using it. And as of November 2026 Google has not documented a feature that reads digitalSourceType from JSON-LD: its image metadata page lists Digital Source Type only as an embedded IPTC field, not as a structured data property. Use it as an accurate description, not as a lever.

For text, the property is a poor fit for light assistance. An article researched with AI help but written and edited by a person is not trained algorithmic media in any useful sense. Reserve the property for works where generation is the main source, and describe lighter assistance in a visible note instead.

Placement matters as much as the value. On a page with a human-written article and one generated illustration, the value belongs on the ImageObject, not on the Article, because it is the image that came from a model. Putting a generative value on the whole Article would describe the text wrongly. Pick the smallest unit the statement is true of, and say the same thing in the caption or disclosure that readers see.

04Who should be the author when AI helped?

The author is the person or organization accountable for the content. Following Google's guidance, keep a human or an organization in the author property and describe AI help in a disclosure, rather than naming a model as the author. If an editor reviewed or fact-checked the piece, say so visibly and name them.

Accurate Person markup supports this. A byline that links to an author page with real credentials tells readers and machines who stands behind the work. Our guide to author and Person schema covers the properties, and the same rule applies throughout: the markup repeats the visible byline, it does not invent one.

• Keep author as the accountable human or organization, never a tool.
• Name the editor or reviewer in the visible page when one checked the piece.
• Put the AI disclosure near the byline or at the end of the article, where readers look for it.
• Use the same wording on every post that shares a workflow, so the disclosure stays consistent across the site.

05How should AI images be labelled?

Images are where machine-readable disclosure is most mature. Google's image metadata documentation lists Digital Source Type among the IPTC photo metadata fields it extracts, and names four values it supports: trainedAlgorithmicMedia, compositeSynthetic, algorithmicMedia and compositeWithTrainedAlgorithmicMedia. The same page explains that IPTC metadata is embedded in the file, so it moves with the image from page to page.

Google's image metadata page also describes how Google can extract C2PA metadata and may show it in the About this image feature, such as whether an image was edited with AI tools, when the signer meets C2PA version and trust list conditions. Two guides in this section cover the details: the IPTC Digital Source Type field for AI images and Content Credentials on WordPress, which matters because WordPress resizing can remove embedded metadata from the copies it serves.

06A disclosure pattern that stays consistent

The pattern below keeps three layers in step: the visible page, the JSON-LD and the image files. It uses only documented properties and makes no claim that any search engine rewards it.

1. Write a short visible disclosure, for example: "Research notes were drafted with an AI tool; this article was written and fact-checked by the named editor."
2. Keep the human or organization author in the Article markup, matching the byline.
3. For a generated or AI-edited image, embed the IPTC Digital Source Type in the file before upload and say so in the caption.
4. Optionally repeat the image's source in JSON-LD with digitalSourceType on the ImageObject, using the same value as the embedded field.
5. Check the served image after upload, because resized copies may not keep embedded metadata.

1{2  "@context": "https://schema.org",3  "@type": "Article",4  "headline": "Example Guide to Spring Planting",5  "author": {16    "@type": "Person",7    "name": "Example Author",8    "url": "https://example.com/authors/example-author/"9  },10  "datePublished": "2026-11-02",11  "dateModified": "2026-11-05",12  "image": {13    "@type": "ImageObject",14    "contentUrl": "https://example.com/uploads/example-garden.jpg",15    "caption": "Illustration created with a generative AI tool.",16    "digitalSourceType": "https://schema.org/TrainedAlgorithmicMediaDigitalSource",217    "creditText": "Example Publisher"18  }19}

A useful disclosure answers three questions in one or two sentences: what the AI tool did, what a person did, and who is accountable for the result. "Drafted with AI" alone leaves readers guessing whether anyone checked the facts. "Outline and first draft produced with an AI tool; reporting, quotes and fact-checking by the named author" tells them. Keep the wording plain, place it where readers expect it, and update it if the workflow changes.

None of this is required for eligibility in any Google feature, and it does not need to be. The goal is honest description: readers see how the content was made, and machines read the same facts. For the wider picture of how structured data for disclosure fits with reviews, bylines and freshness signals, see structured data for disclosure on the handbook home, and browse the authenticity guides for the rest of this section.