Background: Poststroke cognitive impairment (PSCI) is a common complication observed after stroke. Current pharmacologic therapies have no definitive evidence for cognitive recovery or disease progression. Recent studies have verified the positive effect of DL-3-n-butylphthalide (NBP). However, the clinical efficacy and safety are still unclear. The aim of this study was to assess the efficacy of NBP and its harmful effect in the treatment of PSCI. Method: Eligible randomized controlled trials (RCTs) were retrieved from inception to June 2021 from seven medical databases and two clinical registries. The revised Cochrane risk of bias tool (RoB 2.0) was used for methodological quality. RevMan v5.4.1 from Cochrane Collaboration was used for statistical analysis, and Hartung-Knapp-Sidik-Jonkman (HKSJ) method was used for post hoc testing depend on the number of studies. This study has been submitted to PROSPERO with registration number is CRD42021274123. Result: We identified 26 studies with a total sample size of 2,571 patients. The results of this study showed that NBP as monotherapy or combination therapy had better performance in increasing the MoCA (monotherapy: SMDN = 1.05, 95% CI [0.69, 1.42], p < 0.00001; SMDP = 1.06, 95% CI [0.59, 1.52], p < 0.00001. combination: SMDO = 0.81, 95% CI [0.62, 1.01], p < 0.00001; SMDN = 0.90, 95% CI [0.46, 1.33], p < 0.0001; SMDD = 1.04, 95% CI [0.71, 1.38], p < 0.00001), MMSE (monotherapy: MDN = 4.89, 95% CI [4.14, 5.63]), p < 0.00001). combination: SMDO = 1.26, 95% CI [0.97, 1.56], p < 0.00001; SMDC = 1.63, 95% CI [1.28, 1.98], p < 0.00001; SMDN = 2.13, 95% CI [1.52, 2.75], p < 0.00001) and BI (monotherapy: MDN = 13.53, HKSJ 95% CI [9.84, 17.22], p = 0.014. combination: SMDO = 2.24, HKSJ 95%CI [0.37, 4.11], p = 0.032; SMDC = 3.36, 95%CI [2.80, 3.93], p < 0.00001; SMDD = 1.48, 95%CI [1.13, 1.83], p < 0.00001); and decreasing the NIHSS (monotherapy: MDN = −3.86, 95% CI [−5.22, −2.50], p < 0.00001. combination: SMDO = −1.15, 95% CI [−1.31, −0.98], p < 0.00001; SMDC = −1.82, 95% CI [−2.25, −1.40], p < 0.00001) and CSS (combination: MDO = −7.11, 95% CI [−8.42, −5.80], p < 0.00001), with no serious adverse reactions observed. The funnel plot verified the possibility of publication bias. Conclusion: NBP maintains a stable pattern in promoting the recovery of cognitive function and abilities of daily living, as well as reducing the symptoms of neurological deficits. However, there is still a need for more high-quality RCTs to verify its efficacy and safety.