Background We recently identified three distinct spatial‐temporal trajectories of amyloid deposition through the application of a machine‐learning model (i.e., Subtype and Stage Inference SuStaIn) to amyloid positron emission tomography (PET) data. These included Frontal, Parietal, and Occipital subtypes, defined by the earliest regions to show abnormality. The current study aimed to determine whether these subtypes of amyloid accumulation hold value in predicting baseline and longitudinal cognitive functioning in addition to a validated measure of global amyloid burden, i.e. Centiloid (CL) units. Method The Frontal, Parietal, and Occipital subtypes were applied to the amyloid‐PET standard uptake value ratios (SUVr) of 2,510 subjects from 3 cohorts (ADNI, EMIF‐AD, OASIS‐3). Of these, 570 subjects with a high subtype probability assignment (>50%), assigned stage ≥ 1, and available cognition were included in the analyses. Subtypes were compared on global cognition and several cognitive domains including memory, attention, executive functioning, language, and visuospatial functioning. To assess whether subtype assignment added to baseline CL in predicting baseline and longitudinal cognition (range follow‐up=0.4‐11 years), linear models (predictors: subtype and CL) and linear mixed models (predictors: subtype*time and CL*time) were performed, respectively. All models were adjusted for baseline age, sex, education, and baseline cognitive stage. Result Participant characteristics are shown in Table‐1. The majority was assigned to the Frontal subtype (57%), followed by Occipital (22%) and Parietal (20%). After adjusting for baseline CL, the Occipital subtype showed lower baseline language scores than the Parietal subtype (β=‐0.34, p=.046) (Fig‐1A) and lower baseline attention scores than both Frontal (β=‐0.39, p=.047) and Parietal (β=‐0.73, p=.002) subtypes (Fig‐1B, Table‐2). In addition, the Occipital subtype declined faster over time on global cognition (Fig‐2A) and language (Fig‐2B) as compared to the Frontal (βglobal cognition=‐0.35, pglobal cognition=.007; βlanguage=‐0.10, planguage=.015) and Parietal (βglobal cognition=‐0.50, pglobal cognition=.002; βlanguage=‐0.13, planguage=.015) subtypes. Conclusion Our results show the added value of amyloid subtype assignment on top of CL measures on baseline and longitudinal cognition. Specifically, the Occipital subtype was associated with lower baseline performance and showed faster rates of cognitive decline. These findings suggest that subtype assignment based on amyloid PET might improve prognosis and patient stratification in clinical trials.