PurposeMeasuring joint kinematics in the clinic is important for diagnosing injuries, tracking healing and guiding treatments; however, current methods are limited by accuracy and/or feasibility of widespread clinical adoption. Therefore, the purpose of this study was to develop and validate an ultrasound (US)‐based method for measuring knee kinematics during clinical assessments.MethodsWe mimicked four clinical laxity assessments (i.e., anterior, posterior, varus, valgus) on five human cadaver knees using our robotic testing system. We simultaneously collected B‐mode cine loops with an US transducer. We computed the errors in kinematics between those measured using our bone‐tracking algorithm, which cross‐correlated regions of interest across frames of the cine loops, and those measured using optical motion capture with bone pins. Additionally, we conducted studies to determine the effects of loading rate and transducer placement on kinematics measured using our US‐based bone tracking.ResultsPooling the trials at experimental speeds and those downsampled to replicate clinical laxity assessments, the maximum root‐mean‐square errors of knee kinematics using our bone‐tracking algorithm were 2.2 mm and 1.3° for the anterior‐posterior and varus‐valgus laxity assessments, respectively. Repeated laxity assessments proved to have good‐to‐excellent repeatability (intraclass correlation coefficients [ICCs] of 0.81–0.99), but ICCs from repositioning the transducer varied more widely, ranging from poor‐to‐good reproducibility (0.19–0.89).ConclusionOur results demonstrate that US is capable of tracking knee kinematics during dynamic movement. Because US is a safe and commonly used imaging modality, when paired with our bone‐tracking algorithm, US has the potential to assess dynamic knee kinematics across a wide variety of applications in the clinic.Level of EvidenceNot applicable.