Legal Citation Verification

The field studies whether legal citations are real and whether the cited authority supports a particular claim, pinpoint passage, jurisdiction, and time period. Common approaches combine retrieval-augmented generation with citation graphs, source-structure-aware search, and claim-by-claim or multi-agent audits; benchmarks test tasks such as citation retrieval, error detection, and long-form legal research. Results show that models can produce plausible answers while fabricating or misapplying citations, and that finding a relevant case is easier than confirming that a specific page supports a proposition. A central challenge is balancing strict verification and calibrated abstention against missed valid citations, especially across jurisdictions, languages, and disputed legal interpretations.