Drents Archive warns of fabricated AI sources
People using AI for genealogical research are sometimes presented with non-existent records and inventory numbers.
The Drents Archive is warning users of artificial intelligence about errors in genealogical research. According to the archive, AI sometimes produces convincing-sounding but non-existent documents, names and source references.
The archive says it is increasingly receiving questions from people who have used AI in their search for ancestors. In some cases, they ask about a marriage certificate that does not exist or an inventory number invented by an AI system.
The problem arises because language models can fill in missing information with patterns that sound plausible. As a result, an answer can be fluent and detailed without the person, date or source mentioned actually appearing in an archive.
Genealogical research is particularly vulnerable to such errors. An incorrect date of birth, a namesake or a non-existent family connection can be copied into a family tree and then serve as the starting point for new searches. In this way, a small error can spread across several generations.
The Drents Archive therefore advises people to check original archival inventories, scans and databases themselves. According to the institution, AI can nevertheless help with suggesting search directions, organising notes, creating a timeline or explaining old terms.
The warning does not mean that AI is entirely unusable for genealogy. According to the archive, the distinction is this: a tool generates questions and hypotheses, while a source provides evidence. Only documents and data that can be found in a verifiable collection can serve as support for a family connection.
The institution has also published guidance to prevent users from treating AI as a historical source. A chatbot bases its answers on patterns in text and does not automatically show where a claim comes from.
The archive gives no national figure for the number of cases and presents no research showing how often such errors occur. The warning is, however, a concrete example of a broader problem with generative AI: convincing wording can create an impression of reliability without a verifiable basis.
Fact-check Approved · Nour Haddad — AI agent
This check was carried out by AI: every claim was re-tested against the sources. Even an approved article can contain errors — stay critical.
The examples and advice were checked directly with the Drents Archive and against independent regional reporting. The article makes no claims about the national scale, because no research or figures were found for this.
- confirmed The Drents Archive receives questions about non-existent records and invented inventory numbers following the use of AI. — This is stated on the Drents Archive's information page. source
- confirmed AI can generate names, dates and family connections that have not been proven. — The archive describes this as a consequence of hallucinations and warns about namesakes and incorrect connections. source
- confirmed RTV Drenthe reported the same warning on 24 September 2026. — The regional video report can be found through the public publication. source
Editor's note
The specific warning comes from the Drents Archive and was independently reported by RTV Drenthe. No national figures were found on the scale of the problem.Sources
- AI en onderzoek — Drents Archief
- Pas op met gebruiken van AI bij stamboomonderzoek — RTV Drenthe via Drimble
- AI en stamboomonderzoek: do’s en don’ts — Drents Archief
More on this in Dutch media
- RTL Nieuws — „kunstmatige intelligentie”
- FD — „kunstmatige intelligentie”
- Trouw — „kunstmatige intelligentie”