AI and transparency

We use AI tools when we produce articles and social media posts. Not secretly, and not for everything. This page explains where they help, which steps a person takes, and who is named as responsible for the text.

Where we use AI

Three areas in which AI tools do work for us:

  • Research and drafting: sorting topics, gathering sources, writing outlines and first drafts, varying titles and descriptions.
  • Images: the cover images of our blog articles and the visuals of our social media posts are usually generated entirely with AI.
  • Social media: the texts for Instagram, Facebook and LinkedIn are produced in the same way as the articles.

Where we do not use AI

Product figures do not come from a language model. How much credit you receive, how a booking works and how large the community is comes from the product itself, not from an estimate. Reviews, member reports and quotes are neither generated nor polished. Photos showing real people from the team are real photos.

What review comes first

No article goes online without human review. Every draft stays unpublished until a person has read and approved it. Editorial responsibility under Section 18(2) of the German Interstate Media Treaty lies with Arno Hoffrichter, Collective Brain GmbH.

We do not claim that a mistake never slips through. If you spot one, write to us. We correct it and note any change of substance in the article.

How we label images

We distinguish three cases:

  • Images generated entirely with AI carry a visible “AI assisted” note on the image and additionally in the alt text.
  • Real photos whose content has been altered with AI are labelled in the same way.
  • Real photos edited only technically, meaning cropping, brightness or cut-out, are not labelled, because the content does not change. Licensed stock photos carry their author credit in the legal notice instead.

In addition to the visible label, we write a machine-readable provenance tag into every AI image following the IPTC standard. Software that processes images can tell directly that the visual came from an image model.

We do not generate AI images depicting a specific real person. People in our visuals are invented figures, not members and not staff.

Instagram, Facebook and LinkedIn

The note appears on those posts as well. It always appears once AI was involved, including when the image is a real photo and only the text came from the machine.

What the AI Act requires

Since 2 August 2026, Article 50 of the AI Act requires among other things that artificially generated or altered images be marked as such, both visibly and machine-readably. AI-assisted texts are exempt where a person reviews them editorially and a named person carries responsibility. That is exactly what this page describes. If the requirements become more specific in practice, we will update this page and note what changed.

Questions about this page or reports of errors: Contact in the legal notice