Maintaining interpersonal relationships is a major challenge in modern communication. When communicators coordinate information content and the relationship between the two parties, they often face a sense of imbalance caused by dimensional misalignment. This paper proposes and verifies a theoretical model to reveal how LLM can intervene in online communication through strategic AI, thereby regulating users’ perception of rapport and message accuracy. We proposed the concept of a “Tell the Truth” plug-in, which aims to help users deal with the multi-dimensional relationships of information in online communication environments through strategic AI intervention. We conducted an online experiment with 97 participants. They completed three rounds of online response tasks based on work situations. We manipulated the intervention strategy of AI in communication (no AI, strategic AI, non-strategic AI). Results showed that AI intervention improved the semantic accuracy of message texts and affected participants’ perception of interpersonal rapport. This effect was mediated by cognitive load and moderated by task difficulty. Quantitative evidence suggests that strategic AI intervention in online communication is a powerful means to maintain interpersonal relationships. This study provides a new perspective for understanding the information processing mechanism of AI-MC. It applies the concepts of human-computer interaction (HCI) and human-human interaction (HHI) to study the dynamics of interpersonal communication under AI intervention. It provides theoretical guidance and practical inspiration for developers and designers.