Introduction Abbreviated Injury Scale has significant advantages over administrative coding systems for trauma analytics as it was developed specifically for injury, provides greater depth of characterisation of injury and has an integrated severity measure. It is used by trauma registries globally as it allows benchmarking between registries and is used to drive quality improvement. However, the consistency of scoring between individuals is not well understood. An audit was undertaken in six tertiary trauma centres in New Zealand to determine variation between AIS coders. Methods Each of six sites was audited by two experienced auditors. A random selection of case was identified in ISS categories 13–24, 25–44 and 45+. The case notes were pulled, and the auditors independently audited the notes,and then compared their results for a consensus result. The consensus result was then compared with the original coders. Results 111 cases were audited. Coding concordance was found in 31% of cases. Of the 69% of cases where discordant coding was observed, the discordance was attributed to incorrect coding (49%), missed injuries (43%) and other reasons (7%). Head and chest body regions were associated with the greatest number, and largest differences in coding scores. The overall mean difference across all cases was an ISS score of 1. Conclusions The overall accuracy of data held in the New Zealand Trauma Registry (NZTR) is suitable for quality improvement and benchmarking purposes, but more work is needed to improve the accuracy of individual cases, particularly those with head/neck and chest injury. Standardised tools to ensure the accuracy of data in a trauma registry is a gap which needs to be addressed to maintain confidence in a contemporary trauma system.