False news detection remains a critical challenge in today’s information world. With fake news spreading everywhere, it’s crucial to find reliable ways to spot it automatically. Graph neural networks (GNNs) are great for this because they can understand how different pieces of information are connected on social media. This paper investigates the efficacy of GNNs with an attention mechanism in identifying false news using the UPFD dataset. The GNN model, augmented with attention, demonstrates superior performance compared to conventional models and deep learning techniques. Leveraging the attention mechanism allows the model to dynamically weigh contextual data, resulting in enhanced accuracy, precision, recall, and F1-score. The study highlights the importance of relational data in bolstering fake news detection systems. Additionally, considerations are made for model optimization, including node preparation and retweet handling, to mitigate overfitting issues. While the attention-enhanced GNN outperforms other models on the UPFD dataset, alternative models such as SAGE show promise on larger datasets like Gossipcop. Future research directions entail optimizing profile and feature mapping to further enhance GNN precision on direct tweet data.