In the context of rapidly disseminating online information, fake news has emerged as a serious threat to information security and public trust. Traditional detection methods based on content analysis or propagation models often face limitations in handling large-scale, heterogeneous, and weakly labeled data. This study proposes an Adaptive Hypergraph-based Fake News Detection model (A-HGFND), which extends the original HGFND architecture by introducing an adaptive hyperedge weighting mechanism that automatically determines the relative importance of different relational connections. Experiments conducted on the Politifact dataset demonstrate that A-HGFND achieves an accuracy of 92.31% and a precision of 96.15%, indicating a strong capability to detect fake news with a low false-positive rate. The model maintains high performance even under conditions of data scarcity and noisy input, while also providing interpretability through attention-based analysis. With approximately 411 thousand trainable parameters, A-HGFND exhibits computational efficiency and scalability for larger datasets. Despite its promising results, the model’s performance still depends on the quality of hyperedge construction and faces challenges from AI-generated fake content. Future work will focus on developing dynamic hyperedges, integrating multimodal data, and reducing model bias to enhance the reliability and applicability of hypergraph-based models in fake news detection.