We aimed to assess (1)‐whether nuclear β‐catenin is a marker of endometrial precancer, and (2)‐the diagnostic accuracy of β‐catenin immunohistochemistry in the differential diagnosis between benign and premalignant endometrial hyperplasia (EH), defining criteria for its use. Electronic databases were searched for studies evaluating β‐catenin immunohistochemistry in normal endometrium (NE), benign and/or premalignant EH, and endometrioid carcinoma (EC). Odds ratio (OR; p < 0.05), sensitivity, specificity, diagnostic OR (DOR), positive and negative likelihood ratios (LR+, LR−) were calculated. Subgroup analyses were based on the classification system used (WHO or EIN) and criteria to define aberrant β‐catenin expression (only nuclear or cytoplasmic/nuclear). Twelve studies with 1510 specimens were included. Nuclear β‐catenin rate significantly increased from NE to benign EH (OR = 26.01; p = 0.0002, only in WHO subgroup), and from benign EH to premalignant EH (OR = 3.89; p = 0.0002; more markedly in EIN subgroup), but not from premalignant EH to EC (OR = 0.78; p = 0.29). Nuclear β‐catenin accuracy was very low in WHO subgroup (sensitivity = 0.40, specificity = 0.76, LR+ = 1.85, LR− = 0.72; DOR = 2.89) and moderate in EIN subgroup (sensitivity = 0.19, specificity = 1.00, LR+ = 14.80, LR− = 0.83; DOR = 18.14). Cytoplasmic/nuclear β‐catenin accuracy was absent in WHO subgroup (sensitivity = 0.45, specificity = 0.54, LR+ = 1.01, LR− = 1.01; DOR = 0.99) and low in EIN subgroup (sensitivity = 0.57, specificity = 0.86, LR+ = 3.63, LR− = 0.51; DOR = 8.30). Considering nuclear expression and using EIN system, β‐catenin immunohistochemistry might be reliable as rule‐in test for diagnosis of endometrial precancer, with perfect specificity and moderate overall accuracy.