I propose a novel method for efficient object search in realistic environments. I formalize object search as a probabilistic inference problem over possible object locations represented by spatial relations (e.g., in and on). The contribution is twofold. First, I identify five priors, each capturing structure inherent to the physical world that is relevant to the search problem. Second, I propose a formalization of the object search problem that leverages these priors effectively. The formalization in form of a probabilistic graphical model is capable of combining the various sources of information into a consistent probability distribution over object locations. The formalization allows to increase the amount of information in the distribution by propagating knowledge about the world and the potential locations. I use this reasoning method to select actions of a searching artificial agent in a simulated environment and demonstrate in experiments that it results in efficient object search even when using noisy world knowledge extracted from the world wide web.