Background. Paroxetine does not show satisfactory therapeutic effect for generalized anxiety disorder (GAD) patients for the first 2-4 weeks of medication. Diazepam is always concurrently used although it has some shortcomings such as physical dependence and withdrawal reactions. In this study, we aimed to identify whether modified Suanzaorentang (MSZRT), a combined Chinese formula including Suanzaorentang (SZRT) and Zhizichitang (ZZCT), could control the anxiety of GAD for the first 4 weeks of paroxetine medication. Methods. 156 GAD patients were randomized to the treatment of paroxetine, paroxetine-diazepam, or paroxetine-MSZRT for 4 weeks. Hamilton Anxiety Scale (HAMA) Test and Self-Rating Anxiety Scale (SAS) Test were determined each week as the evaluation of clinical efficacy. Adverse events (AEs) were also closely observed by performing the Treatment Emergent Symptom Scale (TESS) Test. Results. Both paroxetine-MSZRT and paroxetine-diazepam decreased more HAMA and SAS total scores than paroxetine from weeks 1 to 3. Paroxetine-MSZRT as well as paroxetine-diazepam had an obviously higher onset rate than paroxetine in each week. After 4 weeks' treatment, the overall effectiveness rate in the paroxetine-MSZRT group (90.00%) was obviously higher than those of the paroxetine group (74.42%) but did not significantly differ from the paroxetine-diazepam group (93.88%). Conclusion. MSZRT had the treatment effect for GAD when paroxetine was used for the first 4 weeks.