Abstract: Objectives: Dementia with Lewy bodies (DLB) is a major cause of degenerative dementia, yet the diagnosis is often missed or mistaken for Alzheimer's disease (AD). We assessed whether the revised Addenbrooke's Cognitive Examination (ACE‐R), a brief test for dementia, differentiates DLB from AD. Methods: We first compared baseline ACE‐R performance in 76 individuals with DLB, 40 individuals with AD and 66 healthy controls. We then investigated the diagnostic accuracy of a simple standardised ‘memory/visuospatial’ ratio calculated from the ACE‐R subscores. Finally, as a comparison a logistic regression machine learning algorithm was trained to classify between DLB and AD. Results: Individuals with AD had poorer memory (p = 0.001) and individuals with DLB had poorer visuospatial function (p = 0.005). Receiver operating characteristics curves confirmed that the ACE‐R total score could differentiate dementia from non‐dementia cases with 98% accuracy, but could not discriminate between dementia types (50%, or chance‐level accuracy). However, a ‘memory/visuospatial’ ratio ≥1.1 differentiated DLB from AD with 82% sensitivity, 68% specificity and 77% mean accuracy. The machine learning classifier did not improve the overall diagnostic accuracy (74%) of the simple ACE‐R subscores ratio. Conclusions: The ACE‐R‐based ‘memory/visuospatial’ ratio, but not total score, demonstrates good clinical utility for the differential diagnosis of DLB from AD.