Background Via the Vienna Visuoconstructional Test 3.0 (VVT 3.0) delayed recall we combined the assessment of visuoconstructive abilities and memory and investigated the test’s potential to support diagnostic processes, including staging and the elaboration of a cognitive profile. Methods We retrospectively analysed the data of 368 patients of the Department of Neurology, Medical University of Vienna, between 04/2014 and 10/2020 that had performed the VVT 3.0. Our sample involved 70 healthy controls (HC), 29 patients with subjective cognitive decline (SCD), 154 patients with mild cognitive impairment (MCI) and 115 patients with Alzheimer’s disease (AD). We investigated the differences in the VVT 3.0 scores, as well as the VVT’s ability to differentiate between AD and nonAD by calculating receiver-operating-characteristic (ROC) curves, ideal cut-offs and a logistic regression model. Results Results stated that the VVT 3.0 delayed recall scores were able to differentiate between all diagnostic groups, respectively, except HC-SCD and SCD-MCI. The ROC analyses determined an AUC of 0.890, 95% CI [0.855; 0.925], P < .001, and the ideal cut-off at 29.5 points that maximised sensitivity at 0.896 and specificity at 0.81. The logistic regression model classified 83.4% of AD patients correctly and delivered a significant Cohen’s Kappa of 0.619 (P < .001). Conclusion As the VVT 3.0 revealed satisfactory values of diagnostic accuracy in our sample, it could enrich clinical diagnosing. However, for more clarity about its informative value in other populations, there remains a need for future studies with other samples.