Grounded in Gricean pragmatics and the theory of corrective negation, this work builds a decision-tree framework for diagnosing whether the Mandarin “not X, but Y” construction is contextually appropriate. Commercial models use the construction at 2.5–9.2 times the density of human reference text, much of it unnecessary; reward models and LLM judges favor it, and both supervised fine-tuning and preference optimization increase its frequency.