A novel digital approach to detecting scalable markers of Alzheimer’s dementia (AD) involves the automated analysis of word properties (WP), enabling both disease detection and prediction of cognitive performance and associated brain correlates. Yet, uncertainty persists regarding the clinical value of WP markers, since no study has compared their discriminative power with standard cognitive and neural measures. We recruited 33 patients with AD and 33 healthy controls, who completed verbal fluency tasks, cognitive tests, and MRI/fMRI scans. Separate machine learning classifiers were trained using (i) WP features from the fluency tasks, (ii) MMSE score, (iii) TMT and digit span scores, (iv) MRI measures, and (v) fMRI measures. The best-performing model for each feature set was evaluated based on mean AUC. WP classification performance (AUC = .85) was comparable (p > .3) to MMSE (AUC = .89), TMT/Digit (AUC = .78), and MRI (AUC = .89) features, and superior to fMRI features (AUC = .65, p < .05). The most important WP feature was word frequency, which showed a negative correlation with the volume of right prefrontal regions involved in executive processing. Our speech approach can identify AD with performance comparable to gold-standard measures, driven by features linked to brain regions implicated in cognitive symptoms of AD. Overall, WP analyses seem non-inferior to standard diagnostic measures, highlighting their potential as a scalable and low-cost tool for dementia