Background: Discriminating mild cognitive impairment (MCI) from subjective cognitive decline (SCD), and healthy controls (HC) remains challenging. Digital tools may overcome paper-and-pencil screening’s limitations. Objective: This study evaluated the diagnostic accuracy and usability of CogniPaz (Daniel Blanco), an ecologically oriented, mobile health (mHealth), gamified cognitive assessment tool exploring memory and executive functions, in discriminating MCI (particularly amnestic phenotypes [aMCI]) from SCD, and HC. Methods: This prospective observational study recruited a convenience sample of consecutive patients with cognitive complaints and HC at La Paz University Hospital Cognitive Disorders Unit from April 2025 to January 2026. Demographic and clinical characteristics, and Montreal Cognitive Assessment (MoCA) scores were collected. An evaluator blinded to the MoCA scores oversaw the automated, self-administered CogniPaz assessment. Participants were classified into HC (no cognitive complaints, MoCA score≥26), SCD (cognitive complaints, MoCA score≥26), and MCI (cognitive complaints without functional impairment, MoCA score<26, subdivided into amnestic [aMCI] and nonamnestic phenotypes). Receiver operating characteristic (ROC) analyses using age-adjusted models, the DeLong test, and bootstrap validation evaluated diagnostic accuracy across groups. The correlation between MoCA and CogniPaz scores was assessed using Spearman rank correlation. Participants with available biomarker data underwent exploratory analysis of CogniPaz’s ability to detect biologically confirmed any-cause neurodegeneration (bND). Usability was tested using an ad hoc Likert questionnaire (1‐9 points). Results: The participants (n=84, 57.1% female; median age 72.0, IQR 62.5-77.0 years) included 27 HC, 17 with SCD, 22 with aMCI, and 18 with nonamnestic MCI. The aMCI group was older than the SCD group (age: median 76.5, IQR 70.8-78.8 vs median 59.0, IQR 50.0-69.0 years; Kruskal-Wallis/Dunn, P =.003). CogniPaz scores correlated with MoCA scores (ρ = 0.81, 95% CI 0.75‐0.89; P< .001). CogniPaz showed good concurrent accuracy, independently of age: MCI vs HC (area under the curve [AUC]=0.880, 95% CI 0.798‐0.961), MCI vs SCD (AUC=0.915, 95% CI 0.841‐0.990), aMCI vs HC (AUC=0.910, 95% CI 0.830‐0.990), and aMCI vs SCD (AUC=0.943, 95% CI 0.878‐1.000). At the provisional <28.5 cutoff point, sensitivity ranged from 85.0% (34/40, 95% CI 70.2‐94.3) to 90.9% (20/22, 95% CI 70.8‐98.9) and specificity from 74.1% (20/27, 95% CI 53.7‐88.9) to 82.4% (14/17, 95% CI 56.6‐96.2) across comparisons. CogniPaz suggested capacity to detect bND (AUC=0.810, 95% CI 0.661‐0.958), with potential incremental value over MoCA and age (odds ratio [OR]=0.78, 95% CI 0.60‐0.94; P =.02). Median administration time was 9.39 (IQR 7.53‐11.38) minutes with 100% (84/84) completion. User satisfaction was high (median score 8.4/9, IQR 7.9‐8.6; Cronbach α=0.84). Conclusions: CogniPaz is an innovative, user-friendly mHealth tool that addresses classical cognitive screening tools’ limitations through a brief, multidomain, minimally supervised gamified format, with high patient acceptance. It correlates with MoCA scores, shows good concurrent capacity to differentiate MCI, particularly aMCI, from SCD and HC, and appears promising for identifying bND. Pending external longitudinal validation, these findings support CogniPaz as a candidate for early cognitive screening in primary care.