Testing methodologies to identify and mitigate hallucinations in AI models: verification of factual consistency of responses, validation of cited sources, detection of invented or distorted information. Adversarial prompt techniques, truthfulness benchmarks, and mitigation strategies for generative systems in production.