Objective We aimed to analyse the clinical utility of the Mattis Dementia Rating Scale (MDRS‐2) for early detection of Alzheimer's disease (AD) and amnestic mild cognitive impairment (MCI) in a sample of Spanish older adults. Methods A total of 125 participants (age = 75.12 ± 6.83, years of education =7.08 ± 3.57) were classified in three diagnostic groups: 45 patients with mild AD, 37 with amnestic MCI—single and multiple domain and 43 cognitively healthy controls (HCs). Reliability, criterion validity and diagnostic accuracy of the MDRS‐2 (total and subscales) were analysed. The MDRS‐2 scores, adjusted by socio‐demographic characteristics, were calculated through hierarchical multiple regression analysis. Results The global scale had adequate reliability (α = 0.736) and good criterion validity (r = 0.760, p < .001) with the Mini‐Mental State Examination. The optimal cut‐off point between AD patients and HCs was 124 (sensitivity [Se] = 97% and specificity [Sp] = 95%), whereas 131 (Se = 89%, Sp = 81%) was the optimal cut‐off point between MCI and HCs. An optimal cut‐off point of 123 had good Se (0.97), but poor Sp (0.56) to differentiate AD and MCI groups. The Memory and Initiation/Perseveration subscales had the highest discriminative capacity between the groups. Conclusions The MDRS‐2 is a reliable and valid instrument for the assessment of cognitive impairment in Spanish older adults. In particular, optimal capacity emerged for the detection of early AD and MCI. Copyright © 2017 John Wiley & Sons, Ltd.