Coret Genealogie uses artificial intelligence (AI) in several places. This statement describes where that happens, which technology is used and how you, as a visitor, can tell that you are looking at AI-generated content.
This AI statement was last modified on 2 August 2026.
Since 2 August 2026 the transparency rules of Article 50 of the European AI Act have applied. Anyone using AI to create or modify content must make that clear, so that a reader or viewer knows what they are looking at. The European Commission has drawn up a code of practice to accompany these rules.
The code of practice comes with three icons that anyone may use freely: a basic icon for content in which AI played a part, the Fully AI-Generated icon for content produced entirely by AI, and the Partially AI-Modified icon for existing, human-made content that has been modified using AI.
Coret Genealogie follows this code of practice voluntarily. Wherever it applies, the corresponding icon is shown alongside the content. Below you can read, per website, where and how AI is used.
Read the European Commission's code of practice
The Coret Genealogie websites exist to be found and read, including by language models and AI assistants. Web spiders that collect content to train AI models or to answer user questions are welcome under two conditions:
Web spiders that harvest content on a large scale without identifying themselves risk being blocked at IP or ASN level.
Content behind a login or a subscription is explicitly excluded and may not be collected.
The texts of Open Archieven, Genealogie Online, Familie Archivaris and Coret Genealogie are written in Dutch and machine-translated into English, German and French using DeepL.
Translation happens in two steps: first from Dutch into English, then from English into German and French. In practice this intermediate step yields a better translation than translating directly from Dutch.
Every translation is then reviewed by a human and adjusted where necessary. No new content is created, only another language version of a human-written text; this is why no AI label is needed here.
Genealogie Online can colourise a black-and-white photo automatically. It uses the publicly available DDColor model, which estimates from image recognition which colours belong to the subjects shown.
The model runs on a server owned by Coret Genealogie. The photos are therefore not sent to an external service and are not used to train models.
Colourising does not create a new photo: the original image is always preserved and remains available. The colourised version is a modification of existing, human-made imagery and therefore carries the AI MODIFIED icon. The colours are an estimate made by the model and not a historical source.
Genealogie Online automatically determines what type of image has been added. It uses a self-trained model that runs on a server owned by Coret Genealogie.
The model chooses from nine types:
If the model is not confident enough, nothing is recorded.
This model does not create or alter any content; it only attaches a label to an existing image so that it is easier to search. This is why no AI label is needed here.
The transcriptions on Open Archieven were produced by computer using handwritten text recognition (HTR). Transkribus and Loghi are used for this: models trained to convert old handwriting into searchable text.
The result is AI-generated text and almost always contains errors. A transcription is an aid to finding and reading a document, and explicitly not a replacement for the original document. When in doubt, always consult the scan of the record.
Open Archieven can produce a summary of a transcription in present-day language. A language model from Mistral AI, a European provider, is used for this.
The summary is generated directly in the language in which you are viewing the website. At the same time the model marks the names, places, occupations and dates it recognises in the text, so that you can click through on them.
The model is instructed to stick strictly to the supplied text and to add nothing. Even so, a language model can make mistakes: the summaries are not checked by a human before they are shown.
These texts are entirely AI-generated and are not a source. That is why every summary carries the AI GENERATED icon. The transcription and the scan of the original document always remain available.
Do you have a question or comment about the use of AI by Coret Genealogie? Please get in touch at ai@coret.org.