<p>Air pollution is a life-threatening public health issue, with millions of premature deaths caused by air pollution every year. To reduce health hazards, environmental damage, and financial losses, effective monitoring and management are crucial. Though the air quality prediction and forecasting have been done since the 1980s, there are significant challenges involved due to their complexity and global impact. Apart from manual AQI measurements by government agencies, novel methods like deep learning techniques can be used to predict AQI in both scalable and cost-effective ways. Deep learning techniques can be accessed using large datasets and detection using nonlinear relationships. A hybrid CNN–LSTM model was developed and trained with AQI datasets (2015–2024) to address challenges in spatial and temporal pattern recognition. The model achieved remarkable accuracy, exceeding existing architectures such as LSTM, GRU, CNN, and WLSTM. The proposed hybrid CNN–LSTM model had outstanding predictive performance, shown by an accuracy of 96.8%, a high R² of 0.92, and low error rates (RMSE = 8.66–13.5, MAE = 0.0514) over datasets from 2015 to 2024. These results confirm the model's robustness and generalization ability for AQI prediction. The findings demonstrate the model's effectiveness in real-time air quality monitoring, assisting policymakers and environmental agencies in making educated decisions for sustainable pollution control.</p>