How Prozly handles AI transparency
Prozly drafts content with AI and marks it as such. You review it, you publish it. This page explains what is generated, by whom, and how you can verify it.
Last updated: 1 October 2026
What is AI-generated
- Draft texts: LinkedIn and Instagram posts, blog articles, newsletters, carousel copy, hooks, comment suggestions.
- Images from the Image Studio, pack images and illustrations inside blog articles.
- Summaries and relevance scores of the sources you follow.
What is not
Your own uploads, photos, logos and brand files are never altered by AI unless you ask for it. Prozly removes technical metadata (such as location data) from uploaded photos for your privacy, and leaves AI images untouched so their marks survive.
Which providers
Text is generated with OpenAI models. Images are generated with OpenAI image models and Google Gemini image models. The current model names are listed in the customer documentation.OpenAI's API does not use your inputs or outputs to train its models by default, and we have not opted in. For Google's Gemini API we use the paid tier, which Google states is not used to improve its products. Prozly itself never trains models on your content; your brand voice and knowledge stay inside your workspace.
How outputs are marked
- AI images carry the IPTC Digital Source Type field (trained algorithmic media, or composite with trained algorithmic media once text or a logo has been placed on them) and, where our signing certificate is configured, a C2PA Content Credentials manifest signed by Prozly.
- The provider's own marks (OpenAI and Google both embed a watermark in the pixels, and OpenAI adds Content Credentials) are kept: the untouched original is stored next to every edited image.
- Hosted blog articles show a notice line and a machine-readable tag when they were written or illustrated with AI. Publishers can switch the visible line off in their blog settings.
- Every draft records the model and the time it was generated.
How to verify
Open an image from Prozly in the Content Credentials verifier at contentcredentials.org/verify, in OpenAI's verifier, or in Google's SynthID detector. For a Prozly-signed image the manifest names Prozly as the signer and the generating model as the software agent. If you need the original provider file for an image, ask us from the workspace that produced it.
You review before publishing
Prozly writes drafts. Nothing is published or scheduled without a person in your workspace deciding to do so. Every editor shows that the content was generated with AI, and the chat assistants say so the first time you use them.
What you remain responsible for
Under the EU AI Act, Prozly is the provider of the AI system and you are the deployer of the content you publish. Visible labels on published AI images that show real people, places or events, and on texts published to inform the public, are your call as the publisher. Prozly ships the marks and the defaults; please do not remove them.
Questions about AI at Prozly?
We answer procurement and compliance questions directly.
Write to security@prozly.ai