Cancer-associated fibroblasts (CAFs) are a major contributor to tumor stromal crosstalk in the tumor microenvironment (TME) and boost tumor progression by promoting angiogenesis and lymphangiogenesis. This study aimed to identify prognostic genes associated with CAFs that lead to high morbidity and mortality in ovarian cancer (OC) patients. We performed bioinformatics analysis in 16 multicenter studies (2,742 patients) and identified CAF-associated hub genes using the weighted gene co-expression network analysis (WGCNA). A machine learning methodology was used to identify COL16A1, COL5A2, GREM1, LUM, SRPX, and TIMP3 and construct a prognostic signature. Subsequently, a series of bioinformatics algorithms indicated risk stratification based on the above signature, suggesting that high-risk patients have a worse prognosis, weaker immune response, and lower tumor mutational burden (TMB) status but may be more sensitive to routine chemotherapeutic agents. Finally, we characterized prognostic markers using cell lines, immunohistochemistry, and single-cell sequencing. In conclusion, these results suggest that the CAF-related signature may be a novel pretreatment guide for anti-CAFs, and prognostic markers in CAFs may be potential therapeutic targets to inhibit OC progression.