The MX3D Bridge is the world’s first structure produced using metal additive manufacturing (AM, i.e. 3D printing), deposited through wire and arc AM (WAAM). The long-term behaviour of structures built using metal AM is unknown and therefore requires the development of techniques that enable this investigation, to facilitate future widespread adoption of these novel manufacturing processes. The MX3D Bridge is instrumented with a comprehensive sensor network (accelerometers, displacement gauges, inclinometers, load cells, strain gauges and thermistors) to enable condition monitoring through the implementation of data-driven anomaly detection techniques. The structure was exposed to environmental and operational variability (EOV) over an 8-week commissioning period, which demonstrated that its response is predominantly temperature driven. Towards the end of the data collection period the decorative end swirls on the bridge were removed, which is taken as a proxy for damage on a structure built using WAAM. EOV is known to mask the damage-sensitive features used in anomaly detection techniques, such as those produced through the end swirl detachment. This paper explores the use of the temperature-based measurement interpretation (TB-MI) approach to: (i) predict the thermal response of the bridge using the iterative regression-based thermal response prediction (IRBTRP) methodology; (ii) remove the influence of EOV from damage sensitive features; and (iii) detect an anomaly event (e.g. the end swirl removal). Accurate predictions are obtained for the thermal response of the MX3D Bridge across 70 sensor measurement signals through the IRBTRP methodology, with 1.2% and 6.4% average prediction errors for the training/validation and unseen testing subperiods, respectively. Moving principal component analysis and cointegration are used with both the measured and thermal response corrected sensor signals for anomaly detection. The detection of the simulated damage is shown to be significantly improved through the removal of EOV, with damage being detected earlier, with greater certainty and across a wider range of sensor clusters (i.e. groups of sensors). The TB-MI approach and IRBTRP methodology can therefore be used to accurately predict the thermal response and detect damage on components and structures produced using metal AM, enabling the advantages of novel AM techniques to be realised within the built environment.