Ecological restoration has been increasingly considering biotic interactions. Different restoration strategies usually rely on different composition and abundance of plants with potential impact on the establishment of plant–pollinator interactions. We evaluated the restoration of plant–pollinator interaction networks in young restoration areas in the South Atlantic rainforest, Brazil. We assessed the relative contribution of two restoration strategies (natural regeneration vs. reforestation), geographic distance, plant composition, pollinator composition, abundance of flowers and insects, and plant–pollinator temporal overlap, that is, phenological coupling, to predict the establishment of pairwise interactions. We expected that restoration strategies would indirectly affect the patterns (identity and frequency) of pairwise interactions due to their influence on the processes driving interactions. We sampled monthly pollinators and the plants they visited on six reforestation sites and six natural regeneration sites during 20 months. We surveyed flower abundance in summer. We analyzed the relative contribution of each factor to predict the identity and frequency of pairwise interactions using structural equation modeling. Contrary to expectations, the restoration strategy did not predict interactions, probably because the sites under restoration were surrounded by natural and conserved landscapes. Since we found no effect of restoration strategies on plant composition, abundance, and phenological coupling, the restoration strategy did not predict interactions. Phenological coupling explained more than half of the interaction patterns, representing the best predictor of interactions followed by abundance and plant composition. Therefore, these predictors should be considered to select plant species in restoration projects that encompass interactions and pollination services.