Mortality from cytomegalovirus disease after marrow transplantation can be reduced by treatment with antiviral drugs based on the detection of viremia and organ involvement. We examined autopsy liver specimens to determine the frequency, extent and outcome of cytomegalovirus hepatitis and whether cytomegalovirus hepatitis occurred in the absence of cytomegalovirus disease elsewhere. Autopsy specimens from 50 transplant patients were evaluated for cytomegalovirus‐infected cells, in five groups of 10, according to extent of CMV during life and at autopsy. Liver sections were examined by routine light microscopy, immunohistochemistry and in situ DNA hybridization. Clinical and laboratory data collected during the last 30 days of life were analysed as markers of liver cytomegalovirus infection. Cytomegalovirus‐infected cells were detected in the livers of 10/10 patients with cytomegalovirus infection during life and widespread cytomegalovirus at autopsy; in 3/20 livers from patients with cytomegalovirus infection during life but negative liver cultures at autopsy; and in 1/10 livers from cytomegalovirus‐seropositive patients who had been without other evidence of cytomegalovirus infection. Histology detected a lower density of cytomegalovirus‐bearing cells per unit area of liver, compared to immunohistochemistry and in situ hybridization. No cytomegalovirus‐infected cells were detected in livers from cytomegalovirus‐seronegative controls. No distinctive clinical or laboratory findings correlated with liver cytomegalovirus detection. CMV liver disease is common in allografted patients with disseminated CMV but may rarely be isolated to the liver, best demonstrated with IHC and ISH. Massive hepatic necrosis from CMV was not seen in any autopsy liver in this study.