This prospective study compared ultra-low-dose chest-abdomen-pelvis CT reconstructed with a deep-learning reconstruction (DLR) algorithm with standard-dose CT in 106 patients undergoing cancer follow-up. Ultra-low-dose CT reduced radiation exposure by approximately 71.5%. Although lesion detection remained comparable for thoracic, skeletal, and larger abdominal lesions, image quality and diagnostic confidence were lower, and small hypoattenuating hepatic lesions were less frequently detected. The stronger DLR setting reduced image noise compared with the Smooth/Smoother settings but did not fully compensate for the limitations of ultra-low-dose imaging. Overall, DLR-based ultra-low-dose CT can markedly reduce radiation dose, but its worse image quality and lower detection of small abdominal lesions may limit its routine use in cancer follow-up.