Modeling Meaning in Generative AI: A Corpus-Assisted Discourse Analysis of Coherence, Framing, and Persuasion

The rapid advancement of generative artificial intelligence has transformed language production, enabling systems to generate fluent and contextually relevant text across diverse domains. However, existing research has largely evaluated such outputs in terms of accuracy and fluency, with limited attention to how AI-generated language functions at the level of discourse. This study examines discourse patterns in generative AI through a corpus-assisted Natural Language Processing (NLP) approach, focusing on coherence, framing, and persuasion. A comparative corpus consisting of AI-generated and human-written texts was constructed to identify recurring linguistic and rhetorical patterns. Quantitative analysis was conducted using corpus tools to examine frequency, collocation, lexical bundles, and discourse markers, while qualitative analysis interpreted these patterns in relation to discourse meaning and communicative function. The findings reveal that AI-generated discourse demonstrates high structural coherence through explicit markers and repetition, yet often lacks conceptual depth and variability. Framing is characterized by controlled neutrality, presenting balanced perspectives while limiting ideological diversity. Persuasion operates through consistency and formulaic expressions rather than contextual and affective engagement. In contrast, human-written discourse exhibits greater flexibility, interpretive richness, and rhetorical variation. The study contributes to bridging the gap between computational linguistics and discourse analysis by integrating corpus-assisted methods with interpretive frameworks. It highlights the need to evaluate generative AI beyond surface-level fluency and underscores the importance of discourse-level analysis in understanding how AI-generated language constructs meaning and influences communication.