BOOTH is a lightweight checkpoint layer for AI applications that validates LLM outputs before they reach users, checking for ambiguity, confidence, and factual accuracy against retrieved evidence. Rather than making models larger, it sits between applications and LLM calls to return structured, defensible decisions and catch confidently wrong answers before deployment.