We derive a probabilistic seismic‐hazard model for the Dead Sea region to allow for seismic risk estimation, which will be part of a subsequent study. This hazard model relies as much as possible on data‐driven approaches, utilizing a seismic catalog compiled for the region by integrating data from different sources. We derive seismicity models using two different smoothing approaches and estimate a hypocentral depth distribution from historical observations. We do not include paleoseismological evidence apart from the observed seismicity. Because ground‐motion records are sparse in the region, we formulate the model in the European Macroseismic intensity scale from 1998. However, the collected macroseismic intensity data are still too few to derive a local intensity prediction equation (IPE). Thus, we choose among existing equations derived for different regions and combine them in a logic tree. Here, motivated by Scherbaum et al. (2010), we propose a two‐step approach to select the IPEs for the logic tree. First, we cluster a set of 10 candidate IPEs to identify groups of models that can be considered similar with respect to the distribution of the predicted values for a selection of the explanatory variables (i.e., magnitude and distance), using a k‐mean approach (Steinhaus, 1956). Then, we apply different ranking techniques (Scherbaum et al., 2009; Kale and Akkar, 2013) to identify within each cluster the most suitable model to be included in the logic tree. The resulting hazard models are consistent with existing probabilistic seismic‐hazard models for the region. We estimate a moderate‐to‐high hazard of intensity grade VII–VIII with 10% probability of exceedance within 50 years in close vicinity to the Dead Sea transform fault, the dominant seismogenic structure in this region.