Discourse Analysis of Advertising Language Created by ChatGPT

This study compares advertising language generated by two versions of ChatGPT—ChatGPT 3.5 and ChatGPT 4o—with advertising language created by humans. To closely analyze the linguistic features of ChatGPT‑produced advertisements, Johnston’s (2008) discourse analysis framework was applied. Among the elements of discourse analysis, particular emphasis was placed on cohesion, repetition, framing, intertextuality, and identity. The advertising language generated by ChatGPT is characterized by weak cohesion, low intertextuality, simple and linear framing, and a structure that can be interpreted immediately without inference. In contrast, human‑created advertising language demonstrates high cohesion, rich intertextuality, multilayered framing, and an emotionally oriented structure that requires inference and interpretation. These differences stem from the fact that humans possess affect and subjective experience, whereas artificial intelligence produces outcomes based on large language model (LLM) data without emotional capacity. Ultimately, because AI lacks the accumulation of personal human experience, its outputs inevitably exclude human experiential depth, revealing a fundamental creative limitation of ChatGPT.