The invention discloses a knowledge graph guided false news detection method which comprises the following steps: step 1, based on a false news detection data set, constructing a knowledge graph as background knowledge of false news detection, and pre-training a knowledge graph embedding model; step 2, extracting entities in a news text to be detected, and connecting the entities to the knowledgegraph; step 3, based on the news text and the entity, obtaining word level enhanced representation of the news text, and extracting news text word level features based on an attention mechanism; step4, obtaining news text word level representation, and extracting news text word level features based on an attention mechanism; step 5, based on the entity attention model, extracting entity featuresin the news text; and step 6, fusing the word-level features, the character-level features and the entity features of the news text to be detected, and performing authenticity detection on the news text to be detected. According to the method, the knowledge graph is introduced to guide the deep learning model to perform false news detection, so that the model recognition accuracy and generalization performance are improved.