Stillbirth affects one in 200 pregnancies in the UK. Understanding the causes of stillbirth is essential to reducing perinatal mortality. Stillbirth certificates represent a potential source of data on perinatal mortality. We aimed to assess whether the information on stillbirth certificates used in the UK is accurate. A retrospective cross-sectional audit of stillbirth certificates issued in a geographical region of the UK in 2009 was undertaken. Data were recorded from the stillbirth certificate and health records. The cause of death was classified using the ReCoDe system. Two hundred and thirteen stillbirth certificates were issued for stillbirths (feticides for fetal anomaly were excluded). Agreement for the primary factor associated with the stillbirth was fair (Kappa = 0.286). This contrasts with the gestation of stillbirth, which was almost complete agreement (Kappa = 0.883). The majority of stillbirths (58.7%) were classified on the certificate as 'unknown cause'. A proportion of 9.4% of stillbirths were classified as congenital anomaly and 8.0% as placental abruption. Only 0.5% of stillbirth certificates cited fetal growth restriction as a relevant condition contributing to death. A total of 49.6% of 'unexplained' stillbirths were associated with fetal growth restriction on review. Errors were present in 77.9% of certificates, including missing co-morbidities (55.9%) and the wrong cause of death (40.4%). The cause(s) of death is (are) not recorded accurately on the UK medical certificate of stillbirth, and the majority of certificates contain one or more errors. Training is required to improve understanding of the causes of stillbirth and completion of medical certificates. Data recorded directly from medical certificate of stillbirths are not sufficiently reliable for descriptive studies of causation and epidemiology.