Based on the social network analysis method, panel data from 254 cities in China spanning from 2006 to 2019 was utilized. A green innovation network was established using a modified gravity model, which described the overall and individual characteristics of the network. It also analyzed the impact utility and mechanisms of centrality features within the network on carbon emissions. The conclusions showed that： ① From 2006 to 2019, the green innovation level of 254 cities in China presented obvious spatial correlation network characteristics. The overall structure of the network was relatively stable, with a steady increase in the number of related relationships and the overall network density value showing an upward trend. The eastern cities represented by Shenzhen, Suzhou, and Wuxi occupied a central position in the green innovation network. ② The improvement of the central location of the urban network had significantly exacerbated carbon emissions, and this impact had certain heterogeneity in different geographical locations and city sizes. Among them, the improvement of the network center position in the northeast, eastern, and western regions had exacerbated carbon emissions, whereas the impact in the central region was not significant. In large cities, the increase in network center location exacerbated carbon emissions； however, it was not significant in small and medium-sized cities. ③ The mesomeric effect showed that the promotion of the network center location could promote carbon emissions through energy consumption.