Current freshwater biomonitoring with diatoms is based on microscopic examination of the morphology of their silica skeleton. This standardized approach is time consuming and requires a high degree of taxonomic expertise. Metabarcoding combined with high-throughput sequencing (HTS) has great potential for next-generation biomonitoring applications but requires standardization. Molecular inventories are strongly influenced by the DNA extraction method used, but the effect of extraction protocols has not been tested to enable selection of the best DNA extraction method for HTS metabarcoding. We used 5 DNA extraction methods combining various types of cell lysis and DNA purification to extract DNA from 8 pure diatom cultures and 8 samples from streams and lakes with differing water quality. We compared the methods based on: 1) quality and purity of the extracted DNA, 2) community inventories obtained from HTS targeting the ribulose-1, 5-bisphosphate carboxylase (rbcL) barcode, and 3) similarity between molecular and microscopy-based inventories of community composition and the Specific Pollution-sensitivity Index [SPI]. A method based on GenElute™-LPA had higher extraction efficiency than the 4 commercial kits but had the highest polymerase chain reaction inhibition level. All 5 methods were efficient for HTS, and method did not affect operational taxonomic unit richness. We observed variations in the relative abundance of some taxa within Nitzschia, Amphora, Encyonema, Gomphonema, and Navicula between 2 of the 5 methods, but method did not affect global diatom community composition or SPI values. SPI values calculated from microscopy-based inventories and molecular inventories based on all 5 extraction methods were strongly correlated. For convenience purposes (high DNA quantity and low cost), we encourage standardization of HTS diatom biomonitoring based on the SA-Gen method.