Observed arrival · 2026-09-04
The Citation Record
An independent quarterly benchmark testing how accurately legal AI systems cite cases, authorities, and controlling law.
Field notes
The benchmark divides citation failure into six categories, including nonexistent cases, incorrect reporters or pages, misrepresented holdings, overruled authority, and nonbinding jurisdictional sources. Attorneys score standardized queries, while CourtListener bulk data handles mechanical checks; held-out items are rotated between editions. The first edition is marked forthcoming for 2026.Q4, and the publisher says its own Citation Firewall will be tested under the same rubric with the publisher recused from scoring.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
One card from the complete issue