Transparency

How we use AI

We use artificial intelligence when writing our texts and say so openly. This page states exactly where AI is involved, where it explicitly is not, and who stands behind the result. These details are current as of 3 August 2026.

The short answer

Our running texts are produced with AI assistance and are reviewed by a human before publication. Our technical data is not produced with AI but comes from manufacturer and retailer information. Our images are currently, without exception, not AI images. If you want the details, read on.

AI text: where AI writes

The descriptive texts on this site are produced with AI assistance. This covers the guides in the guides section, the explanatory sections on the individual model pages and the introductions to the category pages. As a tool we use language models from the Claude family by Anthropic.

The process is always the same. The basis is our own catalogue database with verified technical data. The AI formulates running text from it; it does not invent the facts itself and is not allowed to draw on its own world knowledge when it comes to movement, case dimensions, crystal, water resistance or price. A human then reads the text, checks the details against the source and decides whether it is published. This editorial responsibility is borne by Seydhan Cakmak, named in the legal notice under § 18 para. 2 MStV.

We label every text produced this way directly on the article with the note AI text. So you do not have to look anything up on this page to recognise where AI was involved.

Where no AI is involved

The technical data for each model comes from manufacturer information and is checked against manufacturer or retailer pages before being added to the catalogue. It is not AI output. If you notice an error there, please write to us via the contact page and we will correct it.

The editorial rating is also not AI output but a human assessment of technology, build quality and value for money within the respective class. It is not based on a lab test, and we never claim that it is. Nor do we invent user reviews, assessment results, availability or current prices. What you see as a price with us is a guide price based on the recommended retail price; the binding price is always stated by the shop.

AI image: we currently use none

TopWatchReviews does not currently use a single AI-generated image. So that you can verify this, here is the origin of every type of image you will come across on the site.

Product photos on the model pages are genuine product images and come from Amazon via the Amazon affiliate programme. The portrait photo on the About us page is a genuine photograph of Seydhan Cakmak. The cover image of the homepage is a licensed stock photo, so it is also a genuine photograph.

The stylised watch graphics on cards and overviews are a special case that deserves an explanation. They are neither photos nor AI images, but are drawn by our own program code: from a model's catalogue data, such as watch type and colours, the script calculates a matching vector graphic. This is classic programming following fixed rules, not a learning system. It is still important to us to point out that these graphics only hint at the real appearance of a model and are no substitute for a product photo.

Our rule if we use AI images in future

Should we one day use AI-generated images, for example for schematic illustrations in guides, we will label them twice. First, visibly in the image itself, with a burned-in notice that remains even if someone copies the image out and shares it elsewhere. Second, in the text next to the image. In addition, we write the label into the metadata of the image file so that it can be read by machines.

For product images, we rule out AI images as a matter of principle. A synthetically generated image of a watch could show details that do not exist on the real product, and that contradicts the purpose of this site.

Legal framework

The transparency obligations of the European AI Regulation under Article 50 have applied since 2 August 2026. For texts, the labelling obligation applies where matters of public interest are being reported on, and it expressly does not apply if the content has undergone human review or editorial control and a person bears editorial responsibility.

For our watch buying guidance, both should apply: it is consumer information and not reporting on matters of public interest, and our texts go through editorial review with a named responsible person. In our assessment, we would therefore not be obliged to label the texts. We do it anyway because we consider it the more honest solution and because you should decide for yourself how to weigh a buying recommendation. This page is a self-disclosure and not legal advice.

Questions or doubts

If you have the impression that a text on this site is wrong, misleading or inaccurate, please get in touch. The fastest way is an email to kontakt@topwatchreviews.de. We check every tip-off and correct errors instead of leaving them standing. More about how we work can be found under About us.

Deutsch