Environmental DNA measurement has been widely applied in organism biomonitoring. Different DNA extraction methods may cause changes in yield and stability, resulting in an inaccurate estimation of eDNA, especially when quantitative measurements are performed. This study focused on the DNA extraction method and compared its yield and stability for stream fish and spiked DNA samples. Samples were collected periodically over a year from river and lake water systems and eDNA was spiked into them. The samples were extracted and compared using three methods: using Buffer-AL for initial lysis with the DNeasy Blood and Tissue Kit (Qiagen); using Buffer-ATL for initial lysis and the microfluidic-channel method (BC method). The method using Buffer-ATL in the DNeasy Blood and Tissue Kit showed better stability and a higher yield than the Buffer-AL method. In addition, the BC method, despite being comparatively simple, performed the extraction stably and with relatively high yields. We showed that differences in DNA extraction methods based on the long-term evaluation of eDNA measurements with various methods may cause alterations in DNA yield and stability.