In this study, we screened the expression stability of six reference genes (18S rRNA, β-actin, GAPDH, EF1a, B2M, and HPRT1) in hybrid yellow catfish (n = 6), considering the SBM levels, sampling time points, and different tissues. Four different statistical programs, BestKeeper, NormFinder, Genorm, and Delta Ct, combined with a method that comprehensively considered all results, were used to evaluate the expression stability of these reference genes systematically. The results showed that SBM levels significantly impacted the expression stability of most of the reference genes studied and that this impact was time-, dose-, and tissue-dependent. The expression stability of these six reference genes varied depending on tissue, sampling time point, and SBM dosage. Additionally, more variations were found among different tissues than among different SBM levels or sampling time points. Due to its high expression, 18S rRNA was excluded from the list of candidate reference genes. β-actin and GAPDH in the liver and β-actin, HPRT1 and EF1a in the intestine were the most stable reference genes when SBM levels were considered. HPRT1, and EF1a in tissues sampled at 2 W and EF1a and β-actin in tissues sampled at 4 and 6 W were proposed as two stable reference genes when different tissues were considered. When the sampling time points were considered, β-actin, EF1a, and HPRT1 were the top three stable reference genes in the intestine. In contrast, β-actin and B2M are the most stable reference genes in the liver. In summary, β-actin, EF1a, and HPRT1 were the more stable reference genes in this study. The stability of reference genes depends on the tissues, sampling time points, and SBM diet levels in hybrid yellow catfish. Therefore, attention should be paid to these factors before selecting suitable reference genes for normalizing the target genes.