Multiple modalities of cognitive stimulation (CS) have been designed and tested in samples of patients with probable Alzheimer’s disease (AD). Despite the substantial inter-study variability, an overall positive impact of CS is reported. This impact has been especially observed in general measures of cognition. The mechanisms by which cognitive exercises would be beneficial for high-order cortical functions are still largely undetermined, however. When CS has been applied to patients with mild cognitive impairment (who are at the prodromal stage of AD) more stringent methodological criteria and designs were used and studies have been of greater clinical and research relevance. At this disease stage, a positive impact of CS has been reported in a range of different cognitive domains, and even at a neuro-computational level by the measurement of test-retest modifications of brain function. The effects of CS in healthy adults have also been studied. This population allows researchers to explore and test specific neural mechanisms possibly underlying the effect of pen-and-paper or computerised exercises. The evidence from these studies and those contributing to a better understanding of the pathophysiology of AD has led to devising forms of CS as preventive and therapeutical measures for neurodegenerative diseases based on novel frameworks of brain structure, function and connectivity. An extensive review of the literature was carried out to clarify whether CS is effective in AD and mild cognitive impairment and, together with the evidence from studies in healthy participants, to identify the relevant mechanisms that might sustain this effectiveness.