This chapter clarifies why artificial intelligence (AI) and generative AI (GAI) are not synonymous and situates Large Language Models (LLMs), exemplified by ChatGPT as a specific, as generative subset of AI. Drawing on the European Commission’s High-Level Expert Group definition of AI, we distinguish systems that analyse environments and act towards goals from generative systems that create new outputs (text, image, audio, video) from prior data. We explain how LLMs generate text probabilistically (next-token prediction) and thus simulate meaningful discourse without human-like understanding. The contribution then integrates findings from two in-press studies on teachers of Italian as a second or foreign language to outline current perceptions and uses of GAI and LLMs, while noting that these results are inevitably context-bound and time-sensitive. The chapter also offers a concise review of educational research on LLM, and it introduces Language Teacher Cognition (LTC) as a research framework to ground related analysis. Subsequent sections present research questions, study design, results, and discussion, followed by brief conclusions and avenues for future inquiry.