When describing plant-animal networks, sampling can be performed using plant- or animal-centred approaches. Importantly, while the method affects the characterisation of network structure, how it may affect estimates of interaction dissimilarity across networks is still unknown. Here, we investigated how sampling affects the characterization of pollination networks and their dissimilarities across habitats in a heterogeneous tropical landscape. We also asked whether plant traits influence the difference in interaction specialization according to sampling. Plant-centred networks reported higher interaction and species dissimilarity in space, mainly due to interaction rewiring, while animal-centred networks showed higher specialization and modularity. Floral type and pollination systems affected how specialization was influenced by the sampling method. Combining animal- and plant-centred approaches returned intermediate values for dissimilarity and network metrics, indicating that complementary methods should be used for a better characterization of interaction networks, especially those including groups with distinct mobilities, such as plant and pollinators.