Screening guidelines for cognitive impairment are subjective and inconsistent, often leading to delayed diagnoses. An objective, simple method is needed. Dual-tasking, combined with brain function measurements, may provide sufficient cognitive cost to differentiate between healthy and cognitively impaired populations. It was hypothesized that low complexity (entropy) of brain activity is associated with cognitive impairment, and that dual-tasking would better differentiate between healthy and cognitively impaired groups over resting. Eighty-three participants were recruited: healthy young adults (HY; MoCA = 27.82 ± 1.63, age = 21.13 ± 2.36), cognitively normal older adults (CN; MoCA = 26.59 ± 1.92, age = 75.89 ± 6.93), and mild cognitively impaired/early-stage Alzheimer’s disease older adults (MCI/AD; MoCA =18.96 ± 5.69, age = 78.62 ± 8.54). Functional near-infrared spectroscopy (fNIRS) measured brain activity during a rest and dual-task involving flexion and serial subtraction. The area under the curve (AUC) of multiple scale entropy outcomes was averaged by functional brain regions: dorsolateral prefrontal cortex (dlPFC), anterior prefrontal cortex (aPFC), front eye fields, motor, visuomotor, and primary sensory. During dual-tasking, the average AUC values across all fNIRS channels were 75.02 ± 12.51, 61.44 ± 14.60 and 42.26 ± 22.89 for HY, CN, and MCI/AD groups, respectively. During dual-tasking, differences between AUC values of CN and MCI/AD groups were significant for the average across all channels, primary sensory, aPFC, and dlPFC regions (p textless 0.05). During rest, only the average across all channels was significantly different between CN and MCI/AD groups (p = 0.02). Findings suggest that dual-tasking may better screen for cognitive impairment using fNIRS compared to other task types, especially in regions associated with pathological and behavioral changes in AD.