Current diagnostic and prognostic tools for prostate cancer (PC) are suboptimal, leading to overdiagnosis and overtreatment. Aberrant promoter hypermethylation of specific genes has been suggested as novel candidate biomarkers for PC that may improve diagnosis and prognosis. We here analyzed ST6GALNAC3 and ZNF660 promoter methylation in prostate tissues, and ST6GALNAC3,ZNF660,CCDC181, and HAPLN3 promoter methylation in liquid biopsies. First, using four independent patient sample sets, including a total of 110 nonmalignant (NM) and 705 PC tissue samples, analyzed by methylation‐specific qPCR or methylation array, we found that hypermethylation of ST6GALNAC3 and ZNF660 was highly cancer‐specific with areas under the curve (AUC) of receiver operating characteristic (ROC) curve analysis of 0.917–0.995 and 0.846–0.903, respectively. Furthermore, ZNF660 hypermethylation was significantly associated with biochemical recurrence in two radical prostatectomy (RP) cohorts of 158 and 392 patients and remained significant also in the subsets of patients with Gleason score ≤7 (univariate Cox regression and log‐rank tests, P < 0.05), suggesting that ZNF660 methylation analysis can potentially help to stratify low‐/intermediate‐grade PCs into indolent vs. more aggressive subtypes. Notably, ZNF660 hypermethylation was also significantly associated with poor overall and PC‐specific survival in the RP cohort (n = 158) with long clinical follow‐up available. Moreover, as proof of principle, we successfully detected highly PC‐specific hypermethylated circulating tumor DNA (ctDNA) for ST6GALNAC3,ZNF660,HAPLN3, and CCDC181 in liquid biopsies (serum) from 27 patients with PC vs. 10 patients with BPH, using droplet digital methylation‐specific PCR analysis. Finally, we generated a three‐gene (ST6GALNAC3/CCDC181/HAPLN3) ctDNA hypermethylation model, which detected PC with 100% specificity and 67% sensitivity. In conclusion, we here for the first time demonstrate diagnostic biomarker potential of ST6GALNAC3 and ZNF660 methylation, as well as prognostic biomarker potential of ZNF660. Furthermore, we show that hypermethylation of four genes can be detected in ctDNA in liquid biopsies (serum) from patients with PC.