Explainable AI (XAI) systems must balance effectiveness with ethical constraints to prevent manipulative user interactions. Current frameworks often neglect how anthropomorphic design features and persuasive explanations may exploit cognitive vulnerabilities, risking unintended manipulation. We propose a formal framework for XAI, that dynamically balances ethical persuasion and explanation transparency. The framework integrates three core components: (1) a user vulnerability model quantifying susceptibility to manipulation, (2) a clarity-transparency trade-off mechanism adapting explanations to cognitive and contextual factors, and (3) time-decayed ethical bounds to prevent long-term manipulation. We validate the framework through a health coaching case study, demonstrating its ability to adjust persuasion strength based on user profiles and ethical constraints. Results confirm the framework’s effectiveness in maintaining transparency, without compromising utility.