Alzheimer's disease (AD) pathology begins years before symptoms emerge, making early detection essential. Eye-tracking offers a rapid, non-invasive means for identifying early cognitive decline through oculomotor disturbances. Our analysis of 71 studies investigated the utility of oculomotor tasks in distinguishing Mild Cognitive Impairment (MCI) and AD from healthy controls. Antisaccade tasks consistently demonstrated high utility, revealing impaired accuracy, longer latencies, and reduced gain in AD and MCI patients. In contrast, prosaccade tasks yielded mixed results, while non-saccadic paradigms showed diminished exploratory behavior in AD but inconsistent patterns in MCI. A critical finding was the overwhelming reliance on clinical rather than biological diagnostic criteria across studies (85%), significantly hindering clinical translation. Widespread methodological inconsistencies and small sample sizes (<50 patients in 72% of studies) further limit the generalizability of current evidence. Therefore, while antisaccade tasks are a promising screening tool, future research must adopt biologically defined cohorts and standardized protocols to validate their clinical potential.