Self-Critique Prompting

Self-critique prompting is a pattern where the model is asked to evaluate its own output against specific criteria, surface weaknesses, and suggest improvements — but deliver the critique as an output, not a rewrite. That distinction is what separates it from Self-Refine, which also revises the original answer. Self-critique is useful as a standalone quality-check step in pipelines where something or someone else — a separate agent, a validator, a human reviewer — will decide what to do with the feedback. It is often paired with the Constitutional AI approach, where the critique criteria are drawn from an explicit set of principles rather than left implicit, giving the critique a stable rubric instead of the model's in-context taste.