Background: Primary liver cancer is the fifth most common cancer world-wide, and the second most common cause of death from cancer, with an estimated 841,100 new cases and 781,500 deaths each year. Hepatocellular carcinoma (HCC) accounts for 60–80% of cases, and cholangiocarcinoma 10–40%. We examined global trends in survival for both these sub-types of liver cancer, by country, age, sex and calendar period. Methods: Data on 1,005,032 adults (aged 15–99 years) diagnosed with a primary, invasive malignant neoplasm of the liver or intrahepatic bile ducts between 1995 and 2009 were provided by 243 population-based cancer registries in 60 countries. Analysis was restricted to patients for whom the diagnosis of a primary malignancy had been confirmed by histological or cytological examination, or assignation of a specific morphology code, and to registries from which survival estimates were considered reliable. We estimated both five-year net survival and conditional five-year net survival, for patients who survived to the first anniversary of diagnosis. Funnel plots were used to examine international variation in survival and variation by age and morphology. Results: Data on 578,740 patients from 187 registries in 36 countries were included after quality control. For patients diagnosed during 2004–2009, the pooled estimate of age-standardised five-year net survival for liver cancer was 14.8% (range, 4.4–23.7%), higher than for patients diagnosed during 1995–2000 (11.0%). Survival for patients diagnosed with HCC during 2004–2009 (pooled estimate 17.4%, range 7.7–25.5%) was higher than for those with cholangiocarcinoma (8.4%, range 3.7–16.0%). Survival for patients diagnosed during 2004–2009 was higher in Canada, Italy, Japan, Taiwan and Korea (21.2–23.7%) than the pooled estimate for patients diagnosed some 10 years earlier (1995–2000; 11.0%). Conditional survival in 2004–2009 was also higher in New Zealand, Canada, Taiwan, Korea, and China (42.0–52.7%) than the pooled estimate for 1995–2000 (33.2%). Conclusions: Survival from primary cancers of the liver has increased, but it remains poor in most countries we have examined. International variation in survival highlights the potential to improve outcomes, but prevention must also remain a priority. There is a need for continued and expanded surveillance of survival, especially in low- and middle-income countries, to assess the impact of interventions in policy and treatment. Greater consistency in registration practice and coding of liver cancer would reduce the variation in data quality and further improve the comparability of survival estimates.