Abstract Purpose: Brain involvement occurs in the majority of patients with metastatic melanoma. The potential of circulating tumor DNA (ctDNA) for surveillance and monitoring systemic therapy response in patients with melanoma brain metastases merits investigation. Experimental Design: This study examined circulating BRAF, NRAS, and c-KIT mutations in patients with melanoma with active brain metastases receiving PD-1 inhibitor–based therapy. Intracranial and extracranial disease volumes were measured using the sum of product of diameters, and response assessment performed using RECIST. Longitudinal plasma samples were analyzed for ctDNA over the first 12 weeks of treatment (threshold 2.5 copies/mL plasma). Results: Of a total of 72 patients, 13 patients had intracranial metastases only and 59 patients had concurrent intracranial and extracranial metastases. ctDNA detectability was 0% and 64%, respectively, and detectability was associated with extracranial disease volume (P < 0.01). Undetectable ctDNA on-therapy was associated with extracranial response (P < 0.01) but not intracranial response. The median overall survival in patients with undetectable (n = 34) versus detectable (n = 38) ctDNA at baseline was 39.2 versus 10.6 months [HR, 0.51; 95% confidence interval (CI), 0.28–0.94; P = 0.03] and on-therapy was 39.2 versus 9.2 months (HR, 0.32; 95% CI, 0.16–0.63; P < 0.01). Conclusions: ctDNA remains a strong prognostic biomarker in patients with melanoma with brain metastases, especially in patients with concurrent extracranial disease. However, ctDNA was not able to detect or monitor intracranial disease activity, and we recommend against using ctDNA as a sole test during surveillance and therapeutic monitoring in patients with melanoma.