Gridded meteorological datasets, particularly precipitation products, provide valuable inputs for rainfall–runoff hydrologic modelling, especially in regions with limited spatiotemporal density of ground-based observations. The current research has two main objectives: i) to assess the accuracy of the latest Climate Hazards Center Infrared Precipitation (CHIRP) dataset with a spatial resolution of 0.05° x 0.05°, against gauged precipitation records from 28 meteorological stations distributed across Serbia covering a period of 42 sequential years (1981–2023); and ii) to identify the most effective bias-correction method for reducing systematic errors in the gridded precipitation dataset. The period 1981–2010 was used for calibration, while 2011–2023 was used for independent validation. To this end, four bias-correction methods were implemented, namely, linear regression, monthly mean scaling, monthly quantile mapping, and variance scaling. Their performance was evaluated using well-established statistical metrics, including the coefficient of correlation (R2), root mean square error (RMSE), percent bias (PBIAS), and Kling–Gupta efficiency (KGE). At the national scale, the results identify Linear Regression having the best overall performance where during the validation period the method produced the lowest RMSE (35.96 mm month⁻¹), together with R² = 0.270, KGE = 0.271 and PBIAS = −8.84%. However, substantial spatial variability was observed in the performance of the correction methods. In northern Serbia, for example, monthly scaling provided the best performance with R² = 0.299, KGE = 0.525, RMSE = 32.57 mm month⁻¹ and PBIAS = −9.06%. Overall, the identified regional patterns provide a practical framework for selecting appropriate correction approaches for different parts of the country and support the operational application of gridded precipitation datasets in hydrological forecasting and water-resources management.