AI transparency
Fair for Fitness uses artificial intelligence (AI) as a tool, transparently and with clear rules. This page explains where AI is used, where it deliberately is not, and who bears responsibility.
AI-generated images
Illustrations on this website, in particular the home-page visual, are created with AI and labelled as such directly on the image. They do not show real gyms or real people; they serve purely as illustrative images.
AI-assisted price collection
Some of the tariff data is extracted with the help of AI (Anthropic Claude) from the publicly visible websites and price lists of the studios. Every record passes automatic plausibility checks; we correct detected errors continuously. Every studio page links to its source including an as-of date. Editorial responsibility rests with us.
Ranking without AI
Effective costs and the Fair Score are calculated using a fixed, disclosed formula (70 % price, 30 % flexibility): no AI, no paid placement. Details in the methodology.
Contract scanner
If you upload a contract photo in the “My account” area, an AI system (Anthropic Claude) reads out the key details. The photo is processed only for this recognition and is not stored permanently. Details in the privacy policy.
Chat assistant
The chat on this website is an AI assistant (clearly labelled as AI). It answers questions solely from our public gym data and methodology texts, has no access to internal or personal data, and does not provide legal or health advice. Answers are AI-generated and without guarantee; chat input is sent to an AI service to generate the answer and is not used for advertising. Please do not enter personal data into the chat.
AI-extracted promo texts
The details in the “Current promos” section are partly extracted automatically by AI from the chains' promo pages and published without individual sign-off. They are therefore labelled as AI-generated directly at the section and are provided without guarantee. The provider's own terms always prevail; every promo links to its source.
Guide articles
Our guide articles are produced with AI assistance. Every text is substantively reviewed before publication and published under our editorial responsibility; the note is repeated below each article. A mere spell check would explicitly not be sufficient, so we verify figures, sources and statements. City, chain and studio pages, by contrast, are not written by an AI: they are computed from the data set and filled with fixed text modules.
Machine-readable marking
We label AI-generated content visibly for people and additionally in machine-readable form. Chat assistant answers carry the marking in the response object, in the HTTP headers and in the page source. The AI home-page image is marked up as structured data following the IPTC standard (trainedAlgorithmicMedia). This lets verification tools and search engines detect what comes from an AI.
Development with AI support
This website is built and maintained with the support of AI development tools. All content, calculation logic and editorial decisions are the responsibility of the operator (see legal notice). Using AI while programming does not require labelling; we mention it anyway.
Legal framework and roles
The AI systems used here fall under Regulation (EU) 2024/1689 (AI Act); its transparency obligations under Article 50 have applied since 2 August 2026. For the chat assistant we are a provider within the meaning of the Regulation because we make it available under our own name; for image, text and data tools we are a deployer. We use none of the practices prohibited by Article 5: no emotion recognition, no biometric categorisation, no facial recognition, no social scoring, no manipulative influencing. None of our systems falls under the high-risk cases of Annex III; nothing here decides on access to education, employment, credit or social benefits. Market surveillance in Germany is carried out by the Bundesnetzagentur (coordination and competence centre for the AI Act).
AI literacy
Article 4 of the AI Act requires that the people operating AI systems also understand them. Fair for Fitness is run by one person who knows the models used, their limits and their typical failure modes; every AI-assisted data path has a documented review step before anything goes online.
Questions or corrections? Write to us at info@fair-for-fitness.de. If you believe we have failed to label something, or labelled it wrongly, please tell us and we will correct it promptly; independently of that, you may complain to the market surveillance authority.
Last updated: August 2026.