Large Language Models have become central to a wide range of natural language processing applications, showcasing impressive capabilities in understanding and generating human language. However, these models are prone to hallucinations outputs that appear plausible but conflict with real-world facts underscoring the ubiquitous nature of this issue. The commonality of this issue raises critical questions about the reliability of LLMs in practical applications, where accuracy and trustworthiness are paramount. In this survey, we provide a comprehensive overview of hallucinations in LLMs, discussing their underlying causes and reviewing current detection and mitigation strategies across critical domains. Additionally, we conduct a comparative analysis of existing surveys on LLM hallucinations, highlighting their contributions and limitations in the literature. By examining the implications of hallucinations for real-world applications, this survey aims to enhance understanding of their impact on the reliability of LLMs.