Background: Prostate adenocarcinomas are initially dependent on the androgen receptor (AR) for growth and proliferation. Androgen synthesis inhibitors and AR antagonists are the most commonly used first line treatments in metastatic prostate cancer, but these agents are not curative and virtually all tumors progress to a castration resistant state often associated with AR amplification (AR_amp) and overexpression. Small molecule AR degraders designed to overcome AR_amp/overexpression and other AR alterations associated with resistance are now in clinical development. In parallel, it is important to identify and validate biomarkers that predict sensitivity and inform treatment selection. Methods: CTC samples collected at progression from 114 men with metastatic castration resistant prostate cancer were stained for CK, DAPI, CD45, and AR. AR cRatio (ratio between a cell’s immunofluorescent [IF] intensity and the median IF intensity across all cells) was quantified for each CTC. Low-pass whole genome sequencing (lpWGS) was performed on candidate CK-positive (CK+) CTCs to generate copy number profiles. Cohen’s kappa was used to evaluate the relationship between AR copy number and IF at the single cell level. Dynamics in high AR expressing (AR_high) CTCs were analyzed separately in a longitudinal cohort of patients (n = 18) from whom CTCs were collected at baseline, early on-treatment (4-12 weeks), and at progression on their 1^st AR signaling inhibitor (1L-ARSI). Results: lpWGS was performed on 726 CK+ CTCs, of which 23% (170) were AR_amp and 27% (195) were AR_high (AR cRatio ≥ 3). Fifty-six percent (96/170) of CTCs with AR_amp were AR_high. Overall, a fair agreement was observed between AR copy number and IF (76.1% agreement; k = 0.37, 95%CI = 0.29 – 0.44). At baseline prior to 1L-ARSI, the median CK+ CTC burden was 7.9 CTC/mL (IQR: 1.3 – 10.5). One or more AR_high CTC was detected in 28% (5/18) of patients, but no patient had more than 0.7 AR_high CTCs per mL of blood. AR_high CTCs were not observed in early on-treatment samples. At progression, the median CTC burden was 2.3 CTC/mL (IQR: 0.9 – 9.45) and, despite lower CTC burden, the emergence of a substantial (≥1 CTC/mL) AR_high CTC subpopulation was observed in 17% (3/18) of patients. In samples with ≥1 detected CTC, the mean AR cRatio was 1.106 and 1.777 at baseline and progression, respectively. Conclusions: AR protein can be measured in CTCs with IF staining and is supported by the agreement with data from lpWGS. AR_high CTC subpopulations can be detected at baseline for 1L-ARSI, but these cells were all cleared from circulation after 4-12 weeks of therapy. In a subset of patients at progression, resistance appears at least partially attributable to selection for AR_high CTC subclones.