An overview of current research:
AI vs. Humans
1. GPT-4 passes the Bar Exam (Katz et al, 2024).
The classic: The "ancient" GPT-4 passes the US Bar Exam. It remains disputed how well the AI actually performed.
https://lnkd.in/d-Tawyzf
2. AI vs. Judges (Posner/Saran, 2025)
A complex case is presented to both experienced judges and an AI system for decision. The AI adheres more strictly to precedents, judges focus more on the defendant's personality.
https://lnkd.in/dFM-nQng
3. AI deployment in Shenzhen (Liu/Li, 2025)
Chinese courts use AI to draft judgment reasoning. Decisions are still made by humans, but the study suggests that the content of decisions is also changing.
https://lnkd.in/d_mmbe7C
4. AI in Legal Education
AI-assisted legal practice (Schwarcz et al., 2025)
Law students working with advanced AI models not only work faster but also produce higher quality work.
https://lnkd.in/dAaMYPq2
5. AI writes law exams (Fan et al., 2025)
LLMs write 340 German and English law exams - and then grade them too. In German, the AIs perform worse.
https://lnkd.in/d5JnuqEX
My review: https://lnkd.in/dtyUzfyz
6. Can AI take over professor office hours? (Ouellette et al., 2025)
A patent law textbook is fed into leading AI systems. The AI answers 185 questions based on this. But many answers are only "acceptable" or even wrong.
https://lnkd.in/d3X9p_YW
7. Public Perception Objection overruled! (Schneiders et al., 2025)
Laypeople prefer AI-generated legal advice over that from lawyers - when they don't know the source.
https://lnkd.in/dUm_qZvv
8. AI instead of state courts (Kieffaber et al., 2025)
Start-up founders present the concept of a fully AI-based arbitration court as an alternative to state courts.
https://lnkd.in/dwD-wQTb
Limitations and Risks
9. AI and critical thinking (Lee et al., 2025)
Study with 319 knowledge workers suggests: Higher trust in AI leads to less critical thinking.
https://lnkd.in/drihAyJA
10. RAG is no panacea (Magesh et al., 2024)
Legal research tools with Retrieval Augmented Generation cannot completely prevent hallucinations.
https://lnkd.in/d4JmC-Hd
11. Large models, small effects (Humlum/Vestergaard, 2025)
Barely any impact of AI chatbots on the labor market and at most modest productivity gains - including in legal professions.
https://lnkd.in/dH6PfY9G
Source:
https://www.linkedin.com/feed/update/urn:li:activity:7338803493190656000/
Dr. Sebastian Dötterl
Judge at the Higher Regional Court of Munich
