Abstract High-grade serous ovarian cancer (HGSOC) is responsible for the largest number of ovarian cancer deaths. The frequent therapy-resistant relapses necessitate a better understanding of mechanisms driving therapy resistance. Therefore, we mapped more than a hundred thousand cells of HGSOC patients in different phases of the disease, using single-cell RNA sequencing. Within patients, we compared chemonaive with chemotreated samples. As such, we were able to create a single-cell atlas of different HGSOC lesions and their treatment. This revealed a high intrapatient concordance between spatially distinct metastases. In addition, we found remarkable baseline differences in transcriptomics of ascitic and solid cancer cells, resulting in a different response to chemotherapy. Moreover, we discovered different robust subtypes of cancer-associated fibroblasts (CAF) in all patients. Besides inflammatory CAFs, vascular CAFs, and matrix CAFs, we identified a new CAF subtype that was characterized by high expression of STAR, TSPAN8, and ALDH1A1 and clearly enriched after chemotherapy. Together, tumor heterogeneity in both cancer and stromal cells contributes to therapy resistance in HGSOC and could form the basis of novel therapeutic strategies that differentiate between ascitic and solid disease. Implications: The newly characterized differences between ascitic and solid cancer cells before and after chemotherapy could inform novel treatment strategies for metastatic HGSOC.