This study investigated reliability of freehand three-dimensional ultrasound (3DUS) measurement of in vivo human Achilles tendon (AT) moment arm. Sixteen healthy adults were scanned on 2 separate occasions by a single investigator. 3DUS scans were performed over the free AT, medial malleolus, and lateral malleolus with the ankle passively positioned in maximal dorsiflexion, mid dorsiflexion, neutral, mid plantar flexion and maximal plantar flexion. 3D reconstructions of the AT, medial malleolus, and lateral malleolus were created from manual segmentation of the ultrasound images and used to geometrically determine the AT moment arm using both a straight (straight ATMA) and curved (curved ATMA) tendon line-of-action. Both methods were reliable within- and between-session (intra-class correlation coefficients > 0.92; coefficient of variation < 2.5 %) and revealed that AT moment arm increased by ∼ 7 mm from maximal dorsiflexion (∼ 41mm) to maximal plantar flexion (∼ 48 mm). Failing to account for tendon curvature led to a small overestimation (< 2 mm) of AT moment arm that was most pronounced in ankle plantar flexion, but was less than the minimal detectable change of the method and could be disregarded.