Pinterest ad disclosure: "modified with AI" or "AI-generated person"

Pinterest labels content "AI modified" using IPTC metadata, classifiers and creator disclosure, and shows the label at the bottom left in close-up view. For ads, you disclose through the ellipsis menu, with wording such as "modified with AI" or "contains an AI-generated person", and appeals are available. Use the person wording only when a generated person actually appears.

What the help page says

The page does not publish the classifiers' criteria. Treat classifier results as something to observe, not predict.

Choosing the wording

The two phrasings describe different things. "Modified with AI" fits an image where AI altered or generated content that is not a person, for example a background, a product scene or an edited photograph. "Contains an AI-generated person" fits an image with a generated human. Picking the wrong one invites a mismatch between your disclosure and what a viewer sees.

You control which case you are in at generation time. With the Image API you choose the subject in the prompt and, where the model supports it, pass a real product photo as input_references so the product stays real and the scene is generated.

A pin decision table

The right column is a reading of the two phrases, not a Pinterest ruling.

If a label is wrong

The best defence against a wrong label is a clear record of what the image is and how it was made.

• Use the appeal route the help page describes.
• Keep the source files and generation records for the pin.
• Check whether your metadata says something you did not intend.
• Fix the metadata at the source, then re-upload.

Metadata and honest labels

Pinterest lists IPTC metadata as one detection route, so a file's metadata can trigger a label even if you did nothing. Check what your export settings write, and make sure the file says what you mean it to say.

Keep screenshots of the ad's disclosure state in the campaign folder. If a label changes or is disputed, you have a dated record.