Soil shear strength is the most fundamental property when designing structures in the ground and should be carefully assessed and understood. Several empirical models were introduced to predict the shear strength of unsaturated soils. However, there is uncertainty regarding the applicability and sensitivity of these prediction models. This paper presents a comprehensive verification study to assess the reliability and validity of the existing theoretical models. The results obtained from the prediction models are compared to measured data using thirty experimental data sets. A performance classification program is also conducted to assess the suitability of the analytical models for different soil types as well as over a wide range of matric suctions, saturation degrees, soil densities, soil plasticity, and clay activity. The impact of each single parameter is clarified by the microstructure studies, which also provides insight into the mechanics of unsaturated soils. The results indicated that the applicability of models is more appropriate for sandy soils rather than for clayey ones. The performance of shear strength models tends to decrease with an increase in matric suction, initial density, plasticity index, and clay activity. It is, therefore, recommended that the shear strength estimation models should be carefully selected depending on the soil type and properties. Besides, the analysed results pointed out that the choice to assume the factor chi in the equation of Bishop equals the saturation degree is only suitable for medium-dense soils with low matric suction. This assumption is particularly not effective for clayey soils, or dense soils with high matric suction.