Submerged macrophyte restoration has been widely used to decrease phytoplankton and combat eutrophication in many temperate shallow lakes. However, very limited studies have been done to explore its potential in decreasing phytoplankton in subtropical shallow lakes. We hypothesize that macrophyte restoration can also decrease phytoplankton and shift community structures in subtropical shallow lakes. In order to test our hypothesis, we consistently investigated submerged macrophytes, nutrients, and phytoplankton for three years in a shallow subtropical lake. Multiple analytical methods were employed to assess the effect of macrophyte restoration on the phytoplankton community. The results showed that the density and biomass of total phytoplankton after restoration were less compared to those before the restoration. During the restoration, species diversity significantly increased every year, but the change in species richness was not significant. Moreover, phytoplankton community structure also transformed greatly. Especially cyanobacteria density gradually decreased until almost disappeared; Chlorophyta density also significantly decreased from 4.6 × 10⁶ cell/L to 1.9 × 10⁶ cell/L. Although changes in the densities of other groups were not significant, their ratios in total phytoplankton significantly increased. The changes in four dominant species were also significant during the restoration, with Microcystis aeruginosa, M. incerta, and Chlamydomonas sp. significantly decreasing, but Chlorella pyrenoidosa significantly increasing. Correlation analysis between phytoplankton and environmental factors (macrophyte and water quality) showed that macrophytes were negatively correlated with total phytoplankton, TN, and COD, but total phytoplankton was positively correlated with TN and COD. These relations indicated that macrophyte restoration might not only directly inhibit phytoplankton growth, but also indirectly decrease phytoplankton by both bottom-up and top-down controls of phytoplankton. Therefore, these results basically proved our hypothesis, and more attention should be focused on this method in future lake management.