Administrative Court of Hesse reveals use of AI in writing judgments

A ruling by the Hessian Administrative Court of Appeal dated June 24, 2026, is currently causing a stir – not because of its content, but because of a detail in the footnotes. The ending "source= chatgpt.com " appears five times in the cited internet addresses . As it turns out, a court had used artificial intelligence to research a residency issue – and inadvertently published the software's tracking traces in the process. For clients who rely on the diligence of German authorities and courts, this case is more than just a footnote. It exemplifies how error-prone AI-supported legal research can be – and why this is particularly problematic for foreign clients. This is relevant for skilled workers , students and companies with high levels of commitment.

What happened at the Hessian Administrative Court?

In the proceedings with file number 3 A 2624/25, the Administrative Court referred the question to the European Court of Justice pursuant to Article 267 TFEU as to whether Ukrainian nationals who initially fled to Georgia and only years later to Germany have a right of residence in the EU retention. Ultimately, this concerns the scope of temporary protection under Section 24 of the Residence Act for displaced persons from the Ukraine – a question with far-reaching implications beyond this specific case. While researching the Georgian legal situation, the judges apparently used ChatGPT as a research tool. The links generated in this process – complete with incriminating tracking information – ended up directly in the court ruling.

What errors did the AI-assisted research contain?

This is precisely where the real problem lies: AI systems like ChatGPT can deliver convincing-sounding but factually incorrect or outdated sources. In this particular case, the tool referenced an article from 2025 to substantiate an alleged Georgian government decree from February 24, 2026—a decree that no one could have known about at the time the article was published. Elsewhere, the AI suggested a link led to a decree from 2015, when in fact it led to an amendment from 2025. The court emphasizes that the factual findings were ultimately correct and supported by further sources. However, the case demonstrates that even courts cannot blindly trust AI-generated sources – a principle that follows anyway from the principle of official investigation under Section 86 of the German Administrative Court Procedure Act (VwGO).

Why does this affect foreign skilled workers, students, and companies?

Anyone who as skilled worker from the USA , from Anyone coming to Germany from Great Britain or another third country relies on the diligence of the immigration authorities in residence permit proceedings. Immigration authorities and – in case of dispute – the administrative courts. Especially in complex cases with an international dimension, such as those concerning the legal situation in third countries, bilateral agreements, or transitional arrangements, AI-supported research will become more common in authorities and courts in the future. For applicants, this means: Anyone preparing an application will benefit from AI-supported research. visa , one residence permit or one Anyone relying solely on freely available AI tools or standard legal tech solutions for family reunification risks making the same mistakes that even a higher court has made – only without the resources for a cross-check that a a court or a specialized law firm can provide this service.

How do you protect yourself from AI errors in your own processes?

The Hessian Administrative Court of Appeal emphasized upon inquiry that the use of AI falls under judicial independence, but requires cross-checking – “which has taken place.” This same principle of cross-checking should also apply to your own application. We have been handling residency and visa applications for international clients for years. Skilled workers , companies and Families and are familiar with the administrative practices of Immigration authorities and foreign missions in detail – information that neither an AI chatbot nor a commentary can fully capture. Where an AI tool provides a plausible-sounding but incorrect source, we examine each legal basis against the actually applicable version and current administrative practice. This applies particularly to procedures involving third countries, in the case of Secondments of international employees or when a procedure ultimately results in a This results in legal proceedings before the administrative court.