Abstract Background Determining the presence of amyloid pathology in vivo is possible by visual read (VR) and quantification. VR is the method of reference for clinically assessing amyloid pathology, and guidelines are based on distinguishing positive scans from Alzheimer’s dementia patients from healthy controls negative scans. More quantitative approaches using cut‐off values also focused on identification of early amyloid pathology in preclinical populations. This study investigated whether a VR‐defined cut‐off would perform similarly to recently proposed Centiloid (CL) cut‐offs. Methods 18 Fflutemetamol PET images of 352 cognitively unimpaired (CU) participants from the ALFA+ and 145 patients from the ADC cohort were included (Table 1). Scans were read by an experienced reader according to the product guidelines, collecting the number of positive regions (0‐5) and the final positive/negative classification. Tracer uptake was quantified with the standard CL pipeline. Using VR as the reference standard, previously proposed CL=12/30 cut‐offs were tested for sensitivity and specificity. In addition, a data‐driven cut‐off was determined using ROC analysis and the Youden Index. Results 152 (30.6%) scans were read as positive. The CL=30 cut‐off resulted in 80.3% sensitivity and 100% specificity, while the CL=12 cut‐off greatly improved the sensitivity to 95.2% and kept a high specificity at 95.9%. Based on the ROC analysis, a CL=16 cut‐off was found, with both excellent sensitivity (93.9%) and specificity (98.2%) (AUC of .99, 95% CI: .989‐.998) (Figure 1). Mean CL values significantly increased with increasing number of visually positive regions ( F =568.10, η 2 =.85, p <.01, Figure 2, Table 2). Number of visually positive regions was related to age in the ALFA+ CU population ( F =9.17, η 2 =.12, p <.01). Conclusion Using VR as reference standard, we found a similar CL cut‐off to that proposed by previous work using independent amyloid biomarkers (post‐mortem and CSF). These results illustrate that CL can be as sensitive as VR to early amyloid pathology. Importantly, these results could (at least partially) be attributed the level of experience of the reader and the inclusion of mainly cognitively unimpaired subjects. In addition, our results suggests that it can be of value to visually capture the extent of amyloid burden in addition to negative/positive classifications.