Seasonal forecasts can provide valuable information for managing extreme events, such as triggering drought early-warning measures or informing reservoir operations. However, since these forecasts can have substantial biases for certain areas, postprocessing raw forecasts can be crucial to providing adequate information. An innovative artificial intelligence (AI) postprocessing method based on fuzzy rule-based systems (FRB) has been applied and compared with alternative procedures used in the Jucar River Basin. For this area, six seasonal forecasting systems from the Copernicus Climate Change Service (C3S) and six variables (precipitation, minimum, average and maximum temperature, solar radiation, and wind speed) were considered. ERA5 was used as a reference dataset. Alternative postprocessing methods are linear scaling (LS) and quantile mapping (QM). For each system, variable, and postprocessing alternative, forecasting skill is measured using the continuous rank probability skill score (CRPSS). The results show that, except for precipitation, the relative performance of these methods does not depend on the forecasting systems but on the variable considered. FRB dominates in maximum and minimum temperature and LS in average temperature, and wind speed. However, LS shows the worst performance at maximum and minimum temperatures, while FRB never produces the least ability. In the case of precipitation, the classification between methods depends on the forecasting systems, with FRB yielding the best skill in the majority of them. According to the results in semiarid Mediterranean basins, FRB provides robust and skillful postprocessing across all variables and forecasting systems.