A Survey on AI Search with Large Language Models

Searching for accurate information is a complex task that requires significant effort. Although search engines have transformed the way we access information, they often struggle to fully comprehend intricate human intentions. Recently, Large Language Models (LLMs) have demonstrated impressive abilities in understanding and generating language. However, LLMs face limitations in acquiring external knowledge and accessing the most current information. AI search has evolved by integrating LLMs into the online search process, enabling it to address complex real-world challenges through comprehensive information retrieval and multi-step reasoning, thereby enhancing our ability to browse and search the web effectively. In recent years, substantial progress has been made in refining AI search. This paper provides an in-depth review of these advancements, focusing on text-based AI search, web browsing agents, multimodal AI search, benchmarks, software, and products. We also examine the limitations of current AI search methods and explore promising future directions. For further details, please visit our website https://github.com/swordlidev/Awesome-AI-Search.