Investigations of computerised cognitive training (CCT) show heterogeneous results in slowing age-related cognitive decline. To comprehensively evaluate the effectiveness of serious games-based CCT, integrating control conditions, neurophysiological and blood-based biomarkers, and subjective measures. In this bi-centric randomised controlled trial with parallel groups, 160 participants (mean age 71.3 years) with cognitive impairment ranging from subjective decline to mild cognitive impairment, were pseudo-randomised to three arms: an intervention group receiving CCT immediately, an active control (watching documentaries) and a waitlist condition, which both started the CCT intervention after the control period. Both active arms entailed a 3-month intervention period comprising a total of 60 at-home sessions (five per week) and weekly on-site group meetings. In the intervention group, this was followed by additional 6 months of CCT, with monthly booster sessions to assess long-term training effects. Behavioural and subjective changes were assessed in 3-month intervals. Biological effects were measured by amyloid blood markers and magnetic resonance imaging obtained before and after training. Adherence to the training protocol was consistently high across groups and time points (4.87 sessions per week). Domain-specific cognitive scores showed no significant interaction between groups and time points. Significant cognitive and subjective improvements were observed after long-term training. Voxel-based morphometry revealed no significant changes in grey matter volume following CCT, nor did amyloid levels moderate its effectiveness. Our study demonstrates no benefits of 3 months of CCT on cognitive or biological outcomes. However, positive effects were observed subjectively and after long-term CCT, warranting the inclusion of CCT in multicomponent interventions.