Importance: Dementia prevention frameworks treat depression as a leading modifiable risk factor, but cohort estimates of the association vary widely; whether this variation reflects how cognitive outcomes are ascertained is unknown. Objective: To examine whether associations between depressive symptom trajectories and incident cognitive impairment vary by outcome ascertainment rigor. Design, Setting, and Participants: This harmonized cohort study included 91,043 adults aged 50 years or older from 4 national longitudinal aging cohorts (CHARLS [China], HRS [US], ELSA [England], and SHARE [16 European countries]); baseline waves spanned 2006 to 2011, with follow-up through 2023. Exposure: Depressive symptom trajectory classes (persistent low, persistent high, incident worsening, remitting) derived using group-based trajectory modeling. Main Outcomes and Measures: Incident cognitive impairment ascertained under a 5-level hierarchy of increasing rigor, from self-report (level 1) to clinical consensus diagnosis (level 5; Harmonized Cognitive Assessment Protocol [HCAP]). Cohort-specific Cox models were pooled by random-effects meta-analysis; meta-regression tested ascertainment rigor as a moderator of heterogeneity. Results: Under nonadjudicated outcomes, the incident-worsening trajectory was associated with incident cognitive impairment in CHARLS (HR, 1.31; 95% CI, 1.06-1.62) and SHARE (HR, 1.99; 95% CI, 1.87-2.13) but not HRS (HR, 1.05; 95% CI, 0.95-1.16) or ELSA (HR, 1.10; 95% CI, 0.95-1.27); the persistent-high trajectory was not significant under any nonadjudicated outcome (HR range, 1.07-1.15). Under clinical consensus diagnosis (n = 3,196), all non-low trajectories were independently associated with mild cognitive impairment or dementia (OR range, 1.66-2.29; dementia-only persistent high: OR, 3.32; 95% CI, 1.85-5.95). Effect sizes increased stepwise with ascertainment rigor (meta-regression slope, 0.314 log HR per level; P < .001; residual I2 = 0%; R2 = 100%; exploratory). Findings were robust to lagged, weighted, and competing-risk analyses; biological aging acceleration did not differ across trajectories. Conclusions and Relevance: The detectability and magnitude of depression-related cognitive risk depend on outcome ascertainment rigor. Prevention trials and burden projections relying on self-reported outcomes may underestimate the benefit of depression treatment.