LINGUISTIC AND STYLISTIC FEATURES OF GENERATIVE AI TEXTS IN THE KAZAKH LANGUAGE (BASED ON OPENAI AND A DOMESTIC PLATFORM)

Abstract

This article presents a comparative analysis of the linguistic and stylistic features of texts generated by generative artificial intelligence (AI) platforms in the Kazakh language. The study focuses on responses produced by OpenAI's ChatGPT model and a domestic AI platform. The generated texts were evaluated according to five key criteria: prompt fulfillment (in terms of volume and completeness), adherence to literary norms (grammar and syntax), influence of calques, stylistic register, and lexical richness. A 0–2 point scale was used to quantify text quality. The results indicate that ChatGPT’s responses demonstrate greater stylistic flexibility, clearer argumentation, and more vivid imagery. In contrast, the domestic platform’s outputs were generally shorter, more templated, and often contained calqued expressions or syntactic inaccuracies. These differences largely stem from the nature of the language corpora used, which are predominantly composed of official or formal texts. The study also explores how well culturally embedded language elements are conveyed, revealing limitations in the AI’s ability to reproduce Kazakh-specific imagery and cultural nuances. The findings highlight the importance of aligning AI-generated texts with linguistic norms and appropriate stylistic registers in the development of Kazakh-language generative AI systems.