The thesis explores how Large Language Models (LLMs), specifically OpenAI’s ChatGPT, can be utilized as a tool to support vocabulary building in language learning. After a short introduction to how Large Language Models work, this task-centric research examines the potential of ChatGPT for generating different types of activities for practicing new vocabulary, and also for creating dictionary entries. It investigates various LLM usage scenarios, including generating texts related to specific topics, CEFR levels, genres and regional variations in English. The findings reveal that ChatGPT can generate relevant and grammatically correct content, but human supervision is needed to ensure accuracy and alignment with specific educational goals. Finally, the thesis concludes that LLMs can serve as supplementary tools in vocabulary teaching, they are useful for offering personalized and dynamic content, but their integration requires careful evaluation and teacher oversight.