Knee orthotic devices are commonly prescribed by physicians and medical practitioners for preventive or therapeutic purposes with the aim of supporting, aligning, or immobilising the joint. However, the evaluation of these devices relies on few biomechanical studies or therapeutic trials and the level of their mechanical action remains unclear. The objectives of this work are to validate a developed experimental testing machine regarding its realism as compared to a standardised human limb by using a FE approach, and then use this machine to characterize the efficiency of different categories of orthoses and under different pathological kinematics and investigate the influence of various design characteristics. It was found that the measured mechanical actions should be corrected to compensate for the rigid design of the test machine. Experimental results showed that the tested orthoses highly differed in their ability to restrain motions, and that the stiffening effects of these devices may be able to compensate for deficient internal structures only under low load. Although results remain to be confronted to clinical evidence, this approach paves the way to a standardised procedure for evaluating knee orthoses and developing new designs.