Authorship, Style, and Semantics: Investigating the Language of AI Generated Narratives

With the advent of artificial intelligence (AI) in literary creation, a new landscape has emerged for linguistic study, specifically the study of fiction text created by AI neural networks and large language models. The paper explores the language of AI-written fiction based on a stylistic and discourse analytic approach, by analyzing some representative texts, such as1 the Road (2018), The Day a Computer Writes a Novel (2016), Pharmako-AI (2020), Sunspring (2016), as well as interactive platforms including AI Dungeon (2019) and Project December (2020).This study uses the concepts of stylistics, corpus linguistics, and pragmatics to highlight the common language problems in AI-generated fiction: lexical repetitiveness, syntactic predictability, incoherent discourse structures, pragmatic problems in dialogue, and inconsistent narrative voice. Although AI produces highly fluent and intertextually mimicking texts, they tend to be shallow in terms of semantics, figurative language, and coherent narrative development as found in texts composed by human authors. The analysis also examines the concept of the by-product of algorithmic generation, that is, parody, absurdity and unintended surrealism. This issues questions of authorship, creativity and linguistic boundaries of fiction, insofar as the author is a nonhuman agent. The paper concludes by suggesting that the fictional productions of AI not only question traditional conventions of the literary authority of their authors but also prompt scholars to rethink the role of language in narrative art in the context of machine authorship.