Background: Ambulatory assessment uses digital technology to capture real-time data on mood, mental state and behaviour. It has the potential to enhance traditional clinical outcome measures, but the practical application of these tools fundamentally depends on their performance. Aims: This systematic review aimed to assess the performance of active and passive ambulatory assessment and mood monitoring outcome measures in non-randomised and randomised studies in bipolar disorder over 3 months or longer. We aimed to evaluate their performance against established clinical measures and through inter-ambulatory assessment comparisons. Methods: Systematic review (PROSPERO: CRD42023396473) of performance of mood monitoring and ambulatory assessment protocols in RCTs and non-randomised studies in bipolar disorder. Identified studies were assessed for risk of bias. Due to the very high heterogeneity in included studies and performance metrics we were not able to aggregate the data via meta-analysis. Results: The review included 42 studies with a combined sample of 7,813 participants. We included 28 distinct ambulatory assessment protocols which reported 487 different smartphone-based performance metrics. The considerable variability and inconsistency across these metrics limited our ability to make definitive comparisons of performance. Overall, some active ambulatory assessment approaches showed good performance when compared with established clinical measures. There was a paucity of data examining the performance of passive ambulatory assessment measures. Most studies were rated as having low to moderate risk of bias. Conclusions: While ambulatory assessment holds significant promise, current evidence fails to establish the validity and reliability of passive ambulatory assessment to measure mood. The substantial methodological variation—particularly in how performance metrics are defined and reported—limits meaningful comparison and replication. Greater consistency in ambulatory assessment design and reporting standards is essential to support reliable evaluation and broader adoption of these behavioural assessment tools.