In law, the gap is not legal knowledge but wrong jurisdiction and wrong citations
The same facts can lead to different conclusions in China, California and Germany; models often answer correctly yet fail to cite the right jurisdiction.
Key numbers
- Benchmark scale: three jurisdictions (China, California, Germany), 55 legal issues across contract, consumer, criminal, family and labour law, 6,149 instances grouped into 385 fact patterns, all reviewed by legal professionals.
- Three tasks: single-jurisdiction reasoning, joint cross-jurisdictional comparison, and fine-grained cross-jurisdictional evaluation, plus a Grounded Joint metric scoring answer correctness and legal-source grounding together.
- Results across representative models: they often answer legal questions correctly but struggle to provide accurate cross-jurisdictional legal citations.
Method and sample
arXiv paper “CrossLex: A Source-Grounded Benchmark for Cross-Jurisdictional Legal Reasoning in Large Language Models” (2026-08-02): a same-fact benchmark built from authoritative legal sources across China, California and Germany, pairing jurisdiction-aligned questions with answers and supporting citations, evaluated across representative LLMs.
What it means for brands
The risk in legal work is citation and jurisdiction: an answer that sounds right with the wrong source is still an incident. For firms and legal service providers, being described correctly by AI depends on credentials, practice scope and jurisdictions being structured and verifiable.
How SourceSight handles it
For regulated projects SourceSight first checks whether AI can retrieve credential and scope pages, then the content factory structures licences, certifications and service scope so there is something verifiable to cite.
Limits
The benchmark covers five legal domains across three jurisdictions, not every jurisdiction; it measures model answer quality, not law-firm acquisition outcomes.
References
Run it on your own project
What the same prompts return across 10 engines is only knowable by running them.
Related research
Cross-category recommendations: +36% add-to-cart on item pages, +15% on cart pages
Recommending general merchandise to grocery shoppers — a milk frother with the milk — has long been hard; LLM retrieval plus real-time cart context moved production A/B metrics.
Industries · 2026-10-07Home services: 22% of homeowners already use AI to find a provider, and 80% of businesses don't know how to be found
The local services funnel is moving upstream: before a homeowner opens a website or calls, AI has already synthesised reviews, reputation and service descriptions into one answer.