Transparency
AI in the editorial process
Posts on wischtig.de are written with the help of generative AI. This page says what for, where the line runs, and who answers for the result.
What AI is used for
- Summarising research and sorting sources
- Writing first drafts and outlines that are then reworked
- Translating the German version into English
- Suggesting titles, teasers and metadata
What AI is not used for here
- Nothing ships without human review. No post is published straight out of a model.
- Figures, quotes, names and legal references are checked against the original source. Anything that cannot be verified is cut rather than left to sound plausible.
- No invented experience reports, testimonials or test results.
- No AI-generated images of people and no synthetic voices.
Who is accountable
The person named in the imprint holds editorial responsibility under § 18 (2) MStV. That responsibility cannot be handed to a model: an error in the text is an editorial error, whichever tool produced it.
Legal framework
Article 50(4) of the AI Act (Regulation (EU) 2024/1689) requires deployers to disclose AI-generated text published to inform the public on matters of public interest. It applies from 2 August 2026. It explicitly does not apply where the content has undergone human review or editorial control and a named person holds editorial responsibility.
Both are true here. We label anyway: nobody should have to infer how a text came about from an exemption clause.
Machine-readable
Every post carries the schema.org property digitalSourceType with the value CompositeWithTrainedAlgorithmicMediaDigitalSource in its structured data — the IPTC category for content combining human work and trained models. Crawlers and archives can read the provenance without parsing the prose.
Found an error?
Factual mistakes, outdated legal references, mistranslated passages: a short email is enough, and the correction goes into the text.