Linguistic, cognitive, and pragmatic analysis of artificial intelligence and human speech: theoretical and practical aspects

This article provides an in-depth theoretical and comparative analysis of the linguistic, cognitive, and pragmatic characteristics of texts generated by artificial intelligence (AI) in comparison to real human speech. It explains, from a scientific perspective, how features such as lexical normalization, semantic neutrality, repetitiveness, stylistic smoothness, discursive uniformity, and semantic safety arise in AI-generated texts due to the use of Transformer architectures, probabilistic language models, and statistical generation mechanisms. Differences between AI-generated texts and human speech are examined using examples from English and Uzbek, with particular attention to human cognitive creativity, context-building ability, cultural implicatures, and emotional intonation. The findings have practical significance for AI-detection systems, media text quality assessment, ensuring academic integrity, and improving the efficiency of translation systems.