With disease-modifying therapies emerging for dementia and related conditions, identifying cognitive decline as early as possible is increasingly important. This prospective, single-center, repeated-measures study evaluates whether VR-based eye-tracking pupillometry can provide a practical, non-invasive biomarker of cognitive impairment and its progression over time. Pupil responses are linked to brain arousal systems relevant to cognitive dysfunction, including the locus coeruleus, which is affected early in Alzheimer's disease. Adults aged 18-80 years will be assigned to one of four cohorts (n=35 per cohort): i) Alzheimer's disease (supported by CSF biomarkers), ii) mild cognitive impairment (MCI) without Alzheimer's Disease, iii) depressive disorder with cognitive impairment, iv) healthy controls. Participants will undergo initial assessments at baseline and follow-up visit after 3 and 6 months. At each visit, pupil responses and behavioral metrics are recorded during a pupillary light reflex paradigm, a resting-state fixation block, a working-memory task (N-back), and a reward task. Pupillometric and behavioral metrics will be compared across cohorts and related to routine neuropsychological measures (MoCA, CERAD) and available clinical biomarkers (CSF markers; blood biomarkers). The primary objective is to determine whether task-evoked pupil response profiles sensitively quantify cognitive impairment, differ between cohorts, and track change over time. The long-term goal is to validate an easy-to-use, outpatient-compatible assessment to support objective characterization and monitoring of cognitive disorders.