Lynch syndrome screening in ovarian carcinoma is controversial. The aim of this study was to assess the frequency of deficient mismatch repair (dMMR) protein in a retrospective cohort enriched for non‐high‐grade serous carcinomas and its association with outcome within histological types.Tissue microarrays representing 612 ovarian carcinomas were tested for mismatch repair proteins (MLH1, PMS2, MSH2, and MSH6) by immunohistochemistry. dMMR was detected in 13.8% of endometrioid and 2.4% of clear cell carcinomas, but not in other histological types. Within endometrioid carcinomas, 11 of 25 dMMR cases showed abnormal MLH1/PMS2, 10 cases showed abnormal MSH2/MSH6, and four cases showed only abnormal MSH6, indicating that at least 7.7% of endometrioid carcinomas have dMMR probably related to Lynch syndrome. The four dMMR clear cell carcinomas showed abnormal MSH2/MSH6 in three cases and only abnormal MSH6 in one case, all probably related to Lynch syndrome. Within endometrioid carcinomas, dMMR was significantly associated with age <50 years, synchronous endometrial endometrioid carcinoma, a higher CA125 level at diagnosis, higher FIGO grade, absence of ARID1A, and at least 20 CD8‐positive intraepithelial lymphocytes per high‐power field, but was not associated with cancer‐specific death. Age <50 years, higher CA125 levels at diagnosis and at least 20 CD8‐positive intraepithelial lymphocytes per high‐power field remained significant after adjustment for multiple testing, but their sensitivity for identifying dMMR remained insufficient.Our data support the policy of histotype‐specific Lynch syndrome screening in ovarian carcinoma confined to endometrioid and clear cell carcinomas.