This study aims to analyse the long-term effects (6 months follow-up) of upper limb Robot-assisted Therapy (RT) compared to Traditional physical Therapy (TT), in subacute stroke patients. Although the literature on upper-limb rehabilitation with robots shows increasing evidence of its effectiveness in stroke survivors, the length of time for which the re-learned motor abilities could be maintained is still understudied. A randomized controlled follow-up study was conducted on 48 subacute stroke patients who performed the upper-limb therapy using a planar end-effector robotic system (Experimental Group-EG) or TT (Control Group-CG). The clinical assessments were collected at T0 (baseline), T1 (end of treatment) and T2 (6 months follow-up): Upper Limb part of Fugl-Meyer assessment (FM-UL), total passive Range Of Motion (pROM), Modified Ashworth Scale Shoulder (MAS-S) and Elbow (MAS-E). At T1, the intra-group analysis showed significant gain of FM-UL in both EG and CG, while significant improvement in MAS-S, MAS-E, and pROM were found in the EG only. At T2, significant increase in MAS-S were revealed only in the CG. In FM-UL, pROM and MAS-E the improvements obtained at the end of treatment seem to be maintained at 6 months follow-up in both groups. The inter-groups analysis of FM-UL values at T1 and T2 demonstrated significant differences in favour of EG. In conclusion, upper limb Robot-assisted Therapy may lead a greater reduction of motor impairment in subacute stroke patients compared to Traditional Therapy. The gains observed at the end of treatment persisted over time. No serious adverse events related to the study occurred.