Abstarct: In the digital era, with the expansion of location-baxsed social networks and the widespread recording of spatial data, a massive volume of information has become available that can serve as the foundation for developing point-of-interest recommender systems. Location-baxsed social networks are platforms that allow users to share experiences and discover new places, with Google Maps being a simple and well-known example. In this context, recommender systems play a key role by analyzing user behavior and preferences to suggest new places and enhance user experience. One of the common approaches in this domain is collaborative filtering, which provides personalized recommendations by leveraging similarities in user behavior. Recent studies have shown that combining data-driven approaches with psychological concepts such as the anchoring effect can enhance recommendation accuracy. In cognitive science, the anchoring effect refers to the tendency of individuals to rely on the first piece of information they receive as a reference for subsequent judgments, and its application in recommender systems can be significantly impactful. In this study, the proposed model is inspired by both the anchoring effect and collaborative filtering. In the collaborative filtering component, geographic similarity between users’ activity areas is considered. For each user, the activity region is modeled as a spatial ellipse, and this representation is used to compute similarity between users baxsed on the overlapping area of their ellipses. Subsequently, the final recommendations are generated by incorporating the distance between candidate points and the anchor point. This approach establishes a suitable balance between improving accuracy and controlling computational costs, thereby enabling efficient use of limited hardware resources. Another advantage of this model is its ability to provide step-by-step visualization and high interpretability, which simplifies the understanding and analysis of system performance for researchers and developers. Experimental results on the Gowalla dataset demonstrate that the proposed method outperforms conventional algorithms in terms of precision and recall while also being deployable on lightweight platforms. Overall, this study highlights that integrating psychological and geographical concepts can serve as the basis for designing a simple, effective, and deployable recommender system with strong potential for practical applications and future research in the field of point-of-interest recommendation.