To realize a fast and reliable thickness measurement of multilayer coatings based on terahertz technology, an adaptive teaching-learning-based optimization algorithm was proposed. In this algorithm, the standard Kent chaotic mapping was improved to increase the initial population diversity. Moreover, the teaching and learning phases were enhanced based on step adjustment optimization and suboptimal individual optimization to achieve improved optimization accuracy and efficiency of algorithm. Then, the proposed algorithm was combined with a theoretical model for measuring multilayer sample thicknesses with terahertz waves. A method for determining the thickness of the coatings was developed. Finally, multilayer coatings were prepared and terahertz nondestructive testing experiments were performed. Results show that the efficiency of the proposed method is twice that of the global optimal algorithm. The thickness, refractive index, and extinction coefficient of the multilayer coatings can be obtained quickly in a single measurement in only ~50 s. The relative error of the measured multilayer coating thickness is within 1.5%, and the maximum standard deviation is no more than 1.7 μm. Based on the terahertz measurement signal, the proposed method can be used to efficiently, accurately, and reliably determine the thickness of multilayer coatings.