Background: Computer-based cognitive training programs have been developed with promising results on the maintenance/improvement of cognitive performance in people with dementia.Objective: The objective was to evaluate the effectiveness of the cognitive rehabilitation program “GRADIOR” in people with mild cognitive impairment and mild dementia.Method: This study was a single-blind multicenter randomized clinical trial. Participants were recruited from hospitals/day centers. The experimental group (EG) and control group (CG) received computer-based cognitive training (CCT) and routine daily care, respectively. Outcome measures at T0: baseline, T1: at 4 months, T2: at 12 months were compared within and between-groups.Results: Significant differences or important effect sizes were detected at the intragroup and intergroup level for most variables, observing a trend of improvement and/or maintenance at 4 months by Visual Reasoning of Cambridge Cognitive Examination (CAMCOG), Digit and Arithmetic of WAIS-III, Semantic Verbal Fluency, Mini-Mental State Exam (MMSE), Trail Making Test (TMT)-A-Mistakes and at 12 months by Visual Reasoning of CAMCOG, Digit Symbol of WAIS-III, TMT-B-mistakes, Visual Memory of Rivermead Behavioural Memory Test, Lexical Verbal Fluency-P, Yesavage’s Geriatric Depression Scale (GDS), TMT-A-time scales whose objective was to evaluate some executive functions and/or the memory. The CG presented a worsening trend for most of the measures towards 12 months. There was also a significant interaction between “time and group” for MMSE (F = 8.971; p = 0.03; η2 = 0.019) and the GDS (F = 3.414; p = 0.04; η2 = 0.041), as well as small effect sizes for TMT-A-time (F = 1.641; p = 0.21; η2 = 0.021) and TMT-A-mistakes (F = 0.908; p = 0.41; η2 = 0.019).Conclusion: CCT with GRADIOR has been proved to benefit cognitive functions (ISRCTN:15742788).