When the Machine Speaks: AI-Generated Language, Relational Authenticity, and the Three-Stage Engagement Model in ELT

Abstract

The rapid proliferation of large language models (LLMs) such as ChatGPT and Gemini has introduced a profound conceptual challenge to English Language Teaching (ELT): can language generated by artificial intelligence qualify as authentic input? Traditional definitions of authenticity, grounded in the assumption that genuine texts are produced by and for native speakers, are fundamentally destabilized by AI-generated language, which is statistically derived from human discourse yet produced by no human author. This paper argues that the native-speaker-origin criterion is theoretically insufficient to resolve this challenge, and proposes instead that a relational, process-oriented model of authenticity offers a more productive analytical framework. Extending a foundational distinction between genuineness and authenticity into the era of generative AI, the paper examines whether AI-generated texts can serve as authentic input when engaged through three stages: interaction with text, decoding and interpretation of meaning, and contextually appropriate response production. The paper further considers the dual role of AI as both a source of linguistic input and a mediational tool supporting learner engagement, arguing that the critical determinant of authenticity is not textual origin but the quality and depth of learner engagement with language in use. Pedagogical implications for task design, teacher preparation, and learner agency in AI-mediated language learning environments are discussed.