Skip to main content
Prozly drafts with AI and says so. This page explains the labels you will see, the metadata inside AI images, and the one setting you can change.

​
The labels in the app

  • “Generated with AI. Review before publishing.” appears once in the draft editor sidebar and in the Image Studio toolbar. It is a reminder, not a warning: every draft is a starting point for you to check and edit.
  • “You are chatting with an AI assistant.” appears the first time you use a chat surface (the editor chat, the Image Studio refine chat, the interview flow). Dismiss it once and it stays hidden in that browser.
Neither label is published anywhere. They are for you and your team.

​
The notice on your blog

A hosted blog article that was written or illustrated with AI shows two things:
  • a short “AI-assisted” line under the title, naming your company as the publisher, and a closing line under the article: “Written with AI assistance and reviewed by [your company]”;
  • a machine-readable tag in the page header that search engines and tools can read.
Articles you wrote yourself, without AI text or AI images, show neither. To switch the visible lines off, go to Blog → Settings → Article page and untick Show “AI-assisted” label. The machine-readable tag stays. Before you switch it off: under the EU AI Act, a notice like this may be required by law when AI-generated text is published to inform the public, and the publisher (you) is responsible for it.

​
Provenance inside AI images

Every AI image Prozly serves carries metadata that identifies it as AI-generated:
  • the IPTC Digital Source Type field, set to “trained algorithmic media” for an untouched AI image and “composite with trained algorithmic media” once text, a logo or a canvas edit has been placed on it;
  • C2PA Content Credentials signed by Prozly, where our signing certificate is configured, naming the generating model.
The provider’s own marks are kept too. OpenAI and Google embed a watermark in the pixels that survives editing, and OpenAI adds its own Content Credentials. Because any edit changes the file, Prozly stores the untouched original next to every edited image. To check an image, open it in the Content Credentials verifier at contentcredentials.org/verify. A Prozly-signed image shows Prozly as the signer and the model as the software agent. Photos and logos you upload are different: Prozly removes technical metadata (such as location data) from them for your privacy and never marks them as AI.

​
What stays with you

Prozly provides the AI system; you publish the content. Visible labels on published AI images that show real people, places or events, and on texts meant to inform the public, are your decision as the publisher. Keep the marks in place and use the platform’s own AI-content disclosure (LinkedIn offers one) where it applies.