Abstract
The widespread adoption of generative artificial intelligence (GenAI) stands among the most significant developments of recent years, bringing about substantial changes in the English lexicon. While existing dictionaries have documented this shift through examples such as the new technological sense of the word ‘hallucinate,’ empirical investigations into the mechanics and speed of such semantic change occurrences remain relatively scarce. To address this research gap, the present corpus-based mixed-methods case study investigated the semantic and collocational evolution of the lemma ‘hallucinate,’ together with its nominal and adjectival forms, across a ten-year period (2016–2026). Using data from the 25-billion-word ‘News on the Web’ (NOW) corpus, triangulated with the 89-billion-word ‘English Trends’ corpus, the research compared patterns of lexical behavior during the pre-AI and post-AI periods. Quantitative analyses demonstrated a statistically significant increase in usage frequency following the diffusion of GenAI in late 2022, accompanied by a substantial shift in the target word’s collocational profile from the medical domain (e.g., schizophrenia) toward technological and epistemic contexts (e.g., LLMs, output). Qualitative concordance analysis revealed that this semantic transition was naturalized in discourse through metalinguistic framing and direct anthropomorphism, reflecting a cognitive tendency to map human psychological conditions onto complex machine operations. The findings provide empirical evidence that semantic change can emerge and stabilize within months, offering critical implications for real-time lexicography and EFL pedagogy.
Keywords: Collocational Profile, Conceptual Metaphor, Corpus Linguistics, Generative AI, Semantic Shift.