In this work, it was investigated how to measure Seebeck coefficient accurately. The offset voltages, between the specimen and measurement wires, might influence the results measured significantly and should be eliminated during measuring process. They do not depend on temperature difference but on temperature and include two parts: the intrinsic component related to the materials and the random one related to the contact. The inversion method could eliminate the offset voltages more accurately than the traditional differential methods, and thus measure Seebeck coefficient more accurately. The accuracy of Seebeck coefficient measurement could be further improved by performing a proper temperature difference, optimizing temperature control, and using an electromagnetic screen. The most accurate results were obtained with a standard deviation of 0.06 μV/K, measured under temperature difference of 1 K, temperature variation of 0.002 K, and with an iron electromagnetic screen.