Hallucination mitigation encompasses prompt-level strategies that reduce a model's tendency to generate plausible but false content. Techniques include grounding instructions that constrain the model to provided sources, retrieval augmentation that supplies factual context, asking the model to express uncertainty, requiring inline citations, and chain-of-verification where the model checks its own claims before finalizing a response.