Fake news is rapidly spreading in cyberspace， posing a serious threat to public health and safety. To effectively address this challenge， researchers have begun exploring fake news detection methods based on machine learning algorithms， aiming to investigate the application and effectiveness of large scale language models in fake news detection. The research objective is to develop a system that can automatically identify and classify the authenticity of news， in order to reduce the spread of false information. Among them， the binary classification method is used to evaluate the authenticity of news， which is divided into two categories： “real news” and “fake news”. In terms of methodology， utilize multiple commonly used LLM platforms and demonstrate how to integrate these tools through API call examples. In addition， function calls are used to search engines， and the RAG （Retrieval Augmented Generation） framework is adopted， which is a method that combines retrieval and generation to assist in generating accurate answers by retrieving external knowledge. Provided a series of examples， including direct inquiry method， source reliability assessment， identification of misleading statements， in depth analysis， citation and reference checks， and comprehensive evaluation， demonstrating how to evaluate the credibility of news. The results indicate that prompt engineering can effectively guide the model to complete specific tasks such as text classification， question answering， and text generation. A large scale language model combining RAG framework and prompt word engineering can significantly improve the accuracy and efficiency of fake news detection.