AI transparency
The German version of this document is authoritative. This translation is provided for understanding; in case of discrepancies the German wording applies. Go to the German version
1. What this page is about
Since 2 August 2026, Article 50 of Regulation (EU) 2024/1689 — the AI Act — has applied. Paragraph 2 requires providers of generative AI systems to mark synthetic outputs in a machine-readable form. Paragraph 4 requires deployers in particular to visibly disclose deepfakes and certain AI-generated or manipulated text published to inform the public on matters of public interest. For clear provenance, this page voluntarily goes further and tells you where artificial intelligence is involved and where it deliberately is not.
A large part of the imagery on this site is generated. We do not hide that and we do not dress it up: we are a small company, we have no film crew, and a film that would never have existed without AI is better to us than no film at all. What we do not do is pass generated material off as evidence.
Every generated image and every generated film on this website carries a visible notice — on the object itself, not only here in the small print.
2. The film, the trailer and the figure
The scroll film “Her own name on the door”
Every image in this film was created with generative AI. The main character is fictional; the story she tells is the experience of many of the people we work with. The notice sits at the top of the page, before the film — not as a mark across the images. Under the AI Act, a work that is evidently a narrative may be labelled in a way that does not interfere with the work itself.
The trailer on the home page
The same material, cut shorter. It plays automatically and without sound, and for most visitors it is their first encounter with generated imagery. That is why the notice stays visible on it.
The thirty-second film on “About us”
Its five shots were built from generated stills — a suitcase in an empty room, a wall of filing cabinets, a bare brick wall, an open window. No person appears in them.
The notice matters especially on that page, and in both directions: directly below the film are eight real photographs from our private archive, taken between 2013 and 2022. Without the distinction, generated material would stand indistinguishably next to evidence — and then one devalues the other.
The digital host
The figure who greets you on this website is an AI-generated moving image of Laddawan — a real person, our co-founder. She does not portray anyone invented, and she is still not a recording. Both are stated on her.
3. Images: what is generated and what is not
Generated and labelled:
- the header images of the pages,
- the industry tiles,
- the cover images of the blog posts,
- the imagery of the film, the trailer and the thirty-second film.
Real photographs, and therefore without a notice:
- our two portraits,
- the images of our own path — private archive, taken between 2013 and 2022,
- the screenshots of the reference projects: real sites, real views.
Also without a notice, but for a different reason: the preview images that appear when a link is shared. They are computed from type, colour and our mark — no image model is involved, and they do not resemble a photograph.
What you will not find here: fabricated evidence. No generated client logos, no invented screen contents pretending to be results, no faces claiming to be us. Where an image would assert a fact it cannot support, there is a real photograph or none.
And where a project is not a client project, we say so: two of our references are marked as concept examples, without invented figures.
4. Text: under editorial responsibility
We use AI tools for research and writing — the way we use a translation memory or a search engine.
Every published text is reviewed for substance before it goes online: for accuracy, for completeness, and for whether it says what we consider correct. A spell check would not be enough — what is meant is a review of the matter itself.
Benjamin Tietz carries editorial responsibility for all content on this website. For posts published under Laddawan’s name there is an additional rule: no text goes live that she has not read and released herself.
We review posts before publication and publish them under our editorial responsibility. Where a text falls within the Article 50(4) disclosure rule for matters of public interest at all, the exception there for human-reviewed content with editorial responsibility therefore applies.
5. Where no AI is at work
One thing on this website looks like artificial intelligence and is not. That deserves disclosure just as much as the opposite case.
The digital host is not an AI system. What she says comes from a fixed set of rules we wrote. She understands nothing, she learns nothing, and nothing you do reaches a language model. Her image is generated — her answers are not.
Should that change, it will be stated here before it goes live, and you will be asked before your first input.
Also not in use: emotion recognition, biometric categorisation, automated assessment of visitors. This website makes no decisions about people.
6. AI literacy under Article 4
Since 2 February 2025, Article 4 of the AI Act has required anyone operating AI systems to have sufficient AI literacy. We operate the systems ourselves.
- We use generative image, video and text tools as well as development assistants — solely for design, research and programming, never for decisions about people.
- The provenance of every delivered image is recorded in a register inside the project; the labelling is derived from it rather than set by hand. A check in the build fails if an entry is missing or points nowhere.
- We are aware of the limits and failure modes of these tools — in particular that they can invent facts. That is precisely why every publication is preceded by a review of substance.
7. When we build for you
If a website we build for you contains AI-generated images, you are the deployer of that website under the AI Act — the visible labelling is then yours to provide. We will build it in on request and tell you which images are affected.
If we build in an AI function and make it available under our own name, we may ourselves become a provider. What that means in a given case, we set out in the quote before anything is built.
8. Status and questions
This statement is dated 1 August 2026. It will be updated as soon as the use of AI on this website changes.
We are glad to answer questions — the ways to reach us are in the legal notice.