The increasingly rapid advances in the technology of artificial intelligence have found tremendous progress in recommender systems, using LLMs as a revolutionary tool to enhance recommendation quality. This literature review focuses on integrating large language models into future recommender systems and describes them in all key stages from feature engineering to scoring and user interaction. The paper highlights the power of these methods to enable context understanding, user customization, and system explainability, transcending problems like data sparsity and cold start. Specific approaches including fine-tuning, prompt engineering, and in-context learning are examined in terms of their efficiency, scalability, and user experience impact. The paper also addresses ethics concerns like fairness and privacy in actual applications. A critical review of recent research and practical applications gives a landscape of the current state of LLM integration in recommender systems and suggests possible entry points for future work to best take advantage of their potentials.