In this article, we develop an analytical framework to study the impact of location-based relay selection strategy on the reliability of cognitive relay networks. By utilizing the tool of stochastic geometry, we first derive a closed-form expression for the reliability-enhanced region (RER), where relaying transmission can achieve higher transmission reliability than direct transmission. Then, we adopt the normalized reliability gain (NRG) to quantify the reliability benefit obtained by using relaying transmission compared to direct transmission, and we obtain the spatial distribution of NRG in the RER. Subsequently, by taking the spatial random nature of relays’ distribution into account, we investigate the reliability benefit obtained by secondary networks with the optimal location-based relay selection (OLB-RS) strategy. To reduce the feedback overhead during relay selection, we propose a region-aware relay selection (RA-RS) strategy and obtain the achievable reliability benefit. The results indicate that the reliability is highly dependent on the location of relay, and the OLB-RS strategy is to select the relay closest to the midpoint between the corresponding secondary source and destination.