In an era of escalating misinformation and emotional polarization, public health discourse has emerged as a contested arena of political communication. This study applies artificial intelligence tools—including topic modeling, lexical profiling, and co-occurrence analysis—to 599 verified texts from the HealthRelease corpus to establish a linguistic and thematic baseline for credible institutional messaging. Results show that public health texts are lexically diverse, syntactically complex, and emotionally neutral, with dominant themes centered on biomedical research and chronic disease prevention. Unlike misinformative content, these documents display low lexical redundancy and minimal semantic blending—traits indicative of epistemic precision rather than affective persuasion. Beyond identifying stylistic differences, the study proposes that AI can formalize the rhetorical “signature” of institutional credibility. This shift in focus—from detecting falsehoods to mapping truth-telling patterns—opens new possibilities for evaluating how public institutions communicate under conditions of distrust, politicization, and algorithmic amplification. The findings contribute to the field of political communication by highlighting how linguistic structure and emotional neutrality can serve as strategic assets in restoring institutional legitimacy. Implications extend beyond health, offering insights for other public domains where truth claims are contested and communicative authority is eroding.