Juro

Jintru is an evidence-grounded AI claim verification workbench designed to address a major problem with existing AI fact-checking tools: they can produce convincing answers without adequately verifying the evidence behind them. Users submit factual claims, and Jintru evaluates them against retrieved web evidence. Instead of relying on a simple true-or-false classification, it uses a five-category verdict system: supported, contradicted, mixed, insufficient evidence, and not verifiable. The system uses bounded, two-stage evidence retrieval to examine both supporting and opposing information. An evidence sufficiency gate prevents the system from generating definitive conclusions when the available evidence is inadequate. It also identifies conflicting sources, validates citations against retrieved excerpts, and explains exactly what remains unknown. Jintru includes safeguards against indirect prompt injection in retrieved web content, source deduplication, independent-source corroboration requirements, and persistent verification history. Its central objective is to make AI-assisted fact-checking more transparent, explainable, and evidence-driven.