Due to global warming, the worldwide retreat of glaciers is causing changes in species diversity, community composition, and species interactions. However, the impact of glacier retreat on interaction diversity and ecological networks remains poorly understood. An integrative understanding of network dynamics may inform conservation actions that support biodiversity and ecosystem functioning after glacier extinction. Here, we address how glacier retreat affects the frequency, diversity, and complexity of plant–pollinator interactions, both directly and indirectly through biodiversity change. We surveyed flower visitors (pollinators) and analyzed pollination networks across a gradient of 170 years of glacier retreat (Mont Miné glacier, Valais, Switzerland) which ranges from patchy grasslands to closed forests. We reported a strong impact of glacier retreat on both plant and pollinator communities. Notably, the diversity of plant–pollinator interactions was sharply affected by glacier retreat: interaction diversity increased few years after glacier retreat, but it ultimately decreased in late stages dominated by forests. In contrast, we found that plant–pollinator network complexity did not change with glacier retreat. Our results indicate that the development of plant–pollinator networks is a two-phases process. In the first phase, glacier retreat makes space to plant colonization. This initial increase in plant diversity drives the increase in pollinator and interaction diversity. The second phase is characterized by turnover as woody species encroaches and dominates the community, decreasing the diversity of plant species in ultimate instance. The local decrease of plant diversity leads to a local decrease in pollinator and interaction diversity. Slowing down woody encroachment and enhancing flower diversity, which is initially supported by the glacial landscape, may be key strategies for halting the erosion of ecological networks while increasing biodiversity and ecosystem functioning. Our research thus can help resolve the overarching question of how to conserve ecosystems once glaciers are extinct, pointing toward a composite role of both habitat structure and biological functions.