The structural properties of network system components often display strikingly similar behavior when probed at a macroscopic perspective. Those structural properties, largely determining their dynamic behavior, are revealed by the mesoscopic structure of the underlying networks. To demonstrate this empirical conclusion, it is necessary to develop a set of tools to accurately quantify the structural similarity between network nodes. In this paper, we propose a method to measure nodes’ similarity based on network communicability. Precisely, the approach takes Jensen-Shannon divergence between the communicability sequences of nodes as the difference measure and then obtains the similarity between nodes. We use some real-world networks and artificial networks as test objects, and evaluate the rationality of the method through the topological structure behavior and dynamical behavior of similar nodes respectively. Interestingly, the similar nodes obtained in our framework have very similar dynamical behaviors, which is crucial because the dynamic behaviors of nodes are highly dependent on the mesoscopic structure of the underlying networks. Furthermore, compared with previous methods, the method presented in this paper can more accurately quantify the similarity between nodes from a global perspective.