Vehicle accidents represent a continual worldwide problem, resulting in considerable fatalities and financial damage. Conventional traffic monitoring systems frequently lack the capability for immediate and precise accident identification, resulting in longer response times. This research introduces an innovative IoT-enabled system for real-time accident detection, integrating Graph Neural Networks (GNNs) with deep learning methodologies, including Convolutional Neural Networks (CNNs) and Long Short-Term Memory networks (LSTMs). The system analyses both spatial and temporal data, guaranteeing precise predictions and minimal lag in reaction times. suggested solution outperforms existing systems, achieving an accuracy of 97.6%, precision of 96.2%, recall of 98.1%, and an F1-score of 97.1%. The hybrid model’s efficiency, especially in varying traffic conditions, illustrates its capacity to manage extensive data and conduct real-time analysis, making it an outstanding option for modern road accident detection. This method not only improves safety but also offers a scalable solution for worldwide application in urban traffic control systems.