Imagine a recommender system that not only knows what you've bought but also *understands* why. That's the promise of Large Language Model Enhanced Recommender Systems (LLMERS). Traditional recommender systems often rely on collaborative filtering—analyzing what similar users have liked—or basic content matching. But they struggle with nuances and can't explain *why* a product is recommended. LLMs, however, offer a deeper level of understanding. This research paper explores how LLMs are being integrated into recommender systems, not as the system itself, but as a powerful enhancement. Three key areas are explored: knowledge enhancement, interaction enhancement, and model enhancement. Knowledge enhancement uses LLMs to generate rich textual descriptions of items and users, filling in the gaps that traditional systems miss. Imagine an LLM summarizing your browsing history to create a detailed preference profile. Interaction enhancement tackles data sparsity, a common recommender system problem. LLMs can generate synthetic interactions, essentially imagining what a user might like based on their profile, thus improving the training data for the core system. Finally, model enhancement injects the LLM's understanding directly into the recommendation model itself, either by initializing the model with pre-trained LLM weights or distilling knowledge from a larger LLM into a smaller, faster recommendation model. This survey highlights a significant trend: the move towards incorporating LLMs into the *training* of recommender systems, rather than using them during live recommendations. This avoids the latency and cost issues associated with real-time LLM inference. The focus is shifting towards fine-tuned, open-source LLMs that are tailored for specific recommendation tasks, from e-commerce to news and even job recommendations. This research underscores the exciting potential of LLMs to transform how we discover products, information, and even opportunities. While challenges remain, such as enhancing user-side understanding and improving explainability, the path towards more intelligent, personalized, and helpful recommendations is becoming clearer.