Background Physical activity is an important part of the diabetes management plan. However, the effects caused by different training durations and styles of Tai Chi have not been evaluated. We conducted an updated systematic review of the effects of Tai Chi on patients with type 2 diabetes based on different training durations and styles. Methods We performed a search for Chinese and English studies in 8 databases. Two reviewers independently selected the eligible trials and conducted a critical appraisal of the methodological quality. Results Seventeen trials were included. Tai Chi was found to have reduced fasting blood glucose (FBG) [SMD = − 0.54, 95% CI (− 0.91, − 0.16), P = 0.005] and HbA1c [SMD = − 0.68, 95% CI (− 1.17, − 0.19), P = 0.006] overall, compared with a control group. Considering the subgroup analysis, the pooled results showed that 24 movements or Yang-style Tai Chi did not significantly reduce FBG after a duration of ≤3 months [SMD = − 0.46, 95% CI (− 1.42, 0.50), P = 0.35] or > 3 months [SMD = − 0.50, 95% CI (− 1.49, 0.49), P = 0.32], nor did it reduce HbA1c [SMD = − 1.22, 95% CI (− 2.90, 0.47), P = 0.16] after a duration > 3 months in all studies. However, other styles of Tai Chi significantly reduced FBG [SMD = − 0.90, 95% CI (− 1.28, − 0.52), P < 0.00001] and HbA1c [SMD = − 0.90, 95% CI (− 1.28, − 0.52), P < 0.00001] after a duration > 3 months, while no significant reduction in FBG [SMD = − 0.34, 95% CI (− 0.76, 0.08), P = 0.12] or HbA1c [SMD = − 0.34, 95% CI (− 0.76, 0.08), P = 0.12] was found after a duration ≤3 months. Conclusions Tai Chi seems to be effective in treating type 2 diabetes. Different training durations and styles result in variable effectiveness. The evidence was insufficient to support whether long-term Tai Chi training was more effective.