Low-carbon innovation plays an essential role in carbon reduction worldwide. This study investigates the nexus between low-carbon innovation, economic growth, and carbon emissions by the dynamic spatial Durbin model from 2007 to 2019, along with environmental policy, energy structure, industrial structure, foreign direct investment, energy intensity, and population density as the control variables. First, the results show that provincial agglomeration of carbon emissions exists. High emission provinces concentrate in major economic zones and energy extraction areas. Second, the effect decomposition results show that long-term and short-term effects are consistent. Low-carbon innovation has a significant mitigation effect on carbon emissions in local regions, which effect, however, is not significant in the adjacent areas. The environmental Kuznets curve hypothesis is validated locally, but all provinces and cities have not reached the inflection point of the environmental Kuznets curve, and the linkage effect in adjacent regions remains insignificant. There is both a direct local abatement effect and a demonstration effect in environmental policy. Third, the heterogeneity test results show that there are synergies between environmental policies and low carbon innovation, and for regulated pilot areas, there is a higher mitigation effect of low carbon innovation. Finally, robustness tests for replacing the spatial weight matrics confirmed the robustness of the previous results. Based on the above findings, policy recommendations include providing targeted financial support to stimulate green Research & Development input, building a regional green technology exchange system to enhance knowledge spillovers, promoting the application of carbon trading policies to more regions and industries, and setting comprehensive development goals for a green and high-quality economy.