The Internet can stimulate social relationships, but it can also reduce the quality of real-world social relationships if over-used. There has been insufficient evidence about the relationship between Internet use and social stress. The aim of this study was to investigate whether social stressors are associated with social Internet use in terms of the reduction and stimulation hypotheses. Using longitudinal data from 430 students over three years, we first conducted a latent class analysis (LCA) to determine students' latent classifications based on three self-reported social stressors at each time point. These three stressors related to intimate relationships, peer relationships, and parent-adolescent relationships. Latent transition analysis (LTA) was then conducted to examine how the latent classifications changed over time, and transition patterns were identified. In the next stage, the relationship between stress transition patterns and social Internet use was assessed. The findings indicated that students could be classified as having ＂high social stress＂ or ＂low social stress＂ at each time point. The transition patterns included ＂all high,＂ ＂all low,＂ and ＂improving.＂ The results showed that students who frequently used the Internet to make new friends experienced high social stress both initially and over time. Increasing Internet use to find boyfriends/girlfriends may have reduced later social stress. However, frequently using the Internet to maintain existing relationships did not affect social stress transitions. In sum, using the Internet for social purposes does not necessarily reduce social stress. Students should employ effective and direct stress coping methods instead of the Internet to release their social stress.