AI Transparency: How We Use AI for Images and Text

TL;DR: Some of our featured images are created with AI — those carry the label “AI-generated image” directly beneath them. A larger share are screenshots of real products, websites and dashboards; those are not AI-generated and deliberately carry no label. Our articles are written with AI assistance and reviewed by a human before publication.

Images: we distinguish two kinds

AI-generated images. For news pieces and explainers we create featured images with AI image models — from schematic illustrations to photorealistic scenes. Every one of them carries the label “AI-generated image” directly below the image, not buried in the footer.

Screenshots. The larger share of our images shows real interfaces: product announcements from OpenAI, Google or Anthropic, dashboards, benchmark tables from papers, n8n workflows, tool UIs. These are not AI-generated — they document what is actually on screen. They therefore carry no AI label. Labelling them as AI-generated would be plainly false and would misrepresent the products shown.

We consider this distinction more important than a convenient blanket label. For a publication about AI, it has to stay visible what is evidence and what is illustration.

Also not AI-generated: our logo, graphics reproduced from original sources, and every figure in our tables and body copy.

Text: AI-assisted, human reviewed

We use AI for research, structuring, translation and drafting — for a publication about AI, anything else would lack credibility. Editorial responsibility is unaffected: every article is reviewed by a human before publication, claims are checked against the original source, and anything that cannot be substantiated is removed or explicitly marked as uncertain. The “Human reviewed” badge on our articles refers to exactly this step.

AI Rockstars is published by Florian Schröder and Ralf Schukay; details on our editorial process are on the About page, and the responsible entity is named in the Imprint. That responsibility cannot be delegated to a model — if something in an article is wrong, it is our mistake, not a software’s.

What we do not use AI for

• Producing benchmarks or measurements. Figures from papers, model cards and vendor documentation are cited, not generated. Where we run our own tests, we say so.
• Recreating screenshots. When we show an interface, it is the real interface — never an AI image that looks like a UI.
• Inventing products, prices or release dates. Language models hallucinate version numbers, features and timelines with great confidence. That is precisely why every product claim is checked against the primary source.
• Generating quotes. Direct quotes come from the linked source.

Legal background

Article 50 of the EU Artificial Intelligence Act applies from 2 August 2026. It requires deployers to disclose AI-generated images that realistically depict existing persons, objects, places or events. We label all AI-generated images, including the clearly schematic ones — partial labelling along a line readers cannot follow would be less transparent, not more.

For AI-assisted text the regulation provides an exemption from the labelling requirement where content undergoes human review and a named person holds editorial responsibility. That is the case here — we state it anyway.

Found a mistake?

If an image is labelled incorrectly or not at all, or a claim does not hold up, tell us — contact details are in the Imprint. We will correct it and note the change in the article.

Last updated: 29 July 2026