Abstract: Liquefaction real-time monitoring and warming techniques are novel ways to mitigate liquefaction hazard. The key point is to establish a reverse liquefaction detection method based on seismic records. In this paper, the reliability analysis of the existing liquefaction detection methods, i.e. Suzuki method, Miyajima method, KY method and SY methods, are presented using same sample data which include as many seismic records on soft sites as possible. Assessing results and improving aspects of the methods are proposed. The analysis indicates: the success detection rates on the collected samples by SY method, Suzuki method, KY method and Miyajima method are 98%, 85%, 77% and 52%, respectively; site category influence on success detecting rates are principally notable; with site softening the success rates by Suzuki method, KY method and Miyajima method decrease and are all under 55% for detecting D sites; a parameter correlated with absolute periods are selected as a detecting index in Suzuki method, Miyajima method and KY method, hence the values on soft sites and liquefied sites cross, resulted in misjudge; selecting a relative change of ground motion parameters as a detecting index in SY method, some details in the method need deep discussing and improving even though its success rates for detecting D sites reach 92%.