The targeting accuracy during CT-guided stereotactic radiofrequency ablation (SRFA) of liver tumours was evaluated in a clinical study. Patients under general anaesthesia were immobilized using a vacuum cushion and respiratory motion control was based on temporary disconnections of the endotracheal tube. An optical-based navigation system was used for 3D trajectory planning and needle placement via a stereotactic aiming device. A control CT with the needles in place was fused with the planning CT for accuracy evaluation. Sub-analysis was performed for "location" (liver segments II-VIII), "approach" (intercostal or transabdominal), "properties" (clear parenchymal, subcapsular, subphrenic, fat, and subphrenic plus fat), and "path length". In 20 patients with 35 liver lesions, a total of 145 needles were placed with mean (±SD) lateral errors of 3.6 ± 2.5 mm at the needle tip, angular errors of 1.3° ± 1.2°, and longitudinal errors at the needle tip of -7.4 ± 6.2 mm. No puncture-related complications were noted. No significant differences of angular errors between segments, approach and properties were recorded. Long paths correlated with smaller angular errors. CT-guided stereotaxy can be considered safe and provided precise multiple needle placement for SRFA of liver lesions at arbitrary trajectories in various segments and locations.