Attorly research: how well our legal AI performs, measured in public
Stable reports on the benchmarks Attorly runs automatically on its production release: what was measured, on how many tasks, over what period, with the limitations and the raw data. The numbers update with each run, failures included.
Statute retrieval benchmark 2026: Norway, Sweden, Denmark and Germany
How often Attorly’s legal research finds the statute that governs a matter, measured every night in four European jurisdictions, with method, limitations and the raw data.
ReadLegal answer accuracy benchmark 2026: hallucination in statute answers
How often Attorly’s assistant gives a wrong or mis-cited legal claim when it answers statute questions for Norway, Sweden, Denmark and Germany, with definitions, method, limitations and every answer.
ReadGerman legal retrieval benchmark 2026: statutes and case law on public datasets
Attorly’s German legal research measured on two third-party datasets, GerLeRB for statutes and GerDaLIR for court decisions, with method, sample, limitations and raw data.
Read