AI transparency
Transparenzhinweis nach Art. 50 KI-Verordnung
Every detail on this page is filled in and each one is a fact we checked — what the system actually stores and for how long, and what our providers' own published terms say. What has not happened is a lawyer reading it end to end. This notice stays until that does, because a policy that reads as settled while being wrong is worse than one that says where it stands.
The EU AI Act's transparency obligations under Article 50 have applied since 2 August 2026. This page states plainly where AI is involved, because the useful version of that obligation is a reader knowing what came from a model and what came from measurement.
What is measured, not generated
Everything numeric in the product is arithmetic on observed data — view counts collected repeatedly over time, compared against each channel's own normal result. No model produces a number you see. Script text is written by a language model and is checked against its cited evidence before it reaches you. Every number and every script can be traced to the videos it came from, which is why the source videos are shown beside them.
Where AI is used
| Step | What the model does |
|---|---|
| Grouping similar videos | An embedding model measures how similar two titles are. Deterministic: the same input gives the same grouping. |
| Naming a topic | A language model writes the short description of what a group of videos has in common. |
| Hooks and scripts | A language model writes the suggested openings and scripts. This output is generated, not measured. |
What that means for you
- Scripts and suggested hooks are AI-generated text. Treat them as a starting point, check them, and make them yours.
- If you publish generated text or media, your own transparency obligations may apply — under the AI Act, and under the platform rules of wherever you publish. YouTube requires disclosure of realistic synthetic content.
- We do not use your data to train models, and neither does the provider. The text is generated by Anthropic's Claude models through the Anthropic API, whose commercial terms state that inputs and outputs from the API are not used to train models. We have not opted into any programme that would change that.
Still to settle
Article 50(2) requires providers of systems generating synthetic content to mark outputs as artificially generated in a machine-readable form. How that applies to text handed to a customer rather than published by us needs a lawyer's view, and the Commission's guidance plus the Code of Practice on AI-generated content should be checked before this page is finalised.