AI-generated content: marking and labels
Two separate levels: technical marking, on the providers' side, and visible information, on the deployers' side.
Machine-readable marking
Providers of systems that generate audio, images, video or text must mark this content in a machine-readable format (§ 2). Provenance metadata, such as the Content Credentials of the C2PA standard, is one way of doing so.
Visible information
Deployers must disclose deepfakes, and generated text published to inform the public on matters of public interest, unless it has had human review and someone holds editorial responsibility (§ 4). A visible label next to the content is the most readable form.
What the scan observes
5 signs of AI tools visible in the page (avatar videos, calls to an AI service from the browser, leftover unedited text…) and 2 provenance or label markers. With no sign of generated content, the criterion does not apply and keeps its 25 points. For C2PA, the scan picks up a mention of the standard or of Content Credentials in the page; it does not open image files.