Background: findings from a recent qualitative study indicate that the perceived clinical relevance of computer-generated STOPP/START recommendations was a key factor affecting their implementation by physician prescribers caring for hospitalised older adults in the SENATOR trial. Aim: to systematically evaluate the clinical relevance of these recommendations and to establish if clinical relevance significantly affected the implementation rate. Methods: a pharmacist–physician pair retrospectively reviewed the case records for all SENATOR trial intervention patients at Cork University Hospital and assigned a degree of clinical relevance for each STOPP/START recommendation based on a previously validated six-point scale. The chi-square test was used to quantify the differences in prescriber implementation rates between recommendations of varying clinical relevance, with statistical significance set at P < 0.05. Results: in 204 intervention patients, the SENATOR software produced 925 STOPP/START recommendations. Nearly three quarters of recommendations were judged to be clinically relevant (73.6%); however, nearly half of these were deemed of ‘possibly low relevance’ (320/681; 47%). Recommendations deemed of higher clinical relevance were significantly more likely to be implemented than those of lower clinical relevance (P < 0.05). Conclusions: a large proportion (61%) of the computer-generated STOPP/START recommendations provided were of potential ‘adverse significance’, of ‘no clinical relevance’ or of ‘possibly low relevance’. The adjudicated clinical relevance of computer-generated medication recommendations significantly affects their implementation. Meticulous software refinement is required for future interventions of this type to increase the proportion of recommendations that are of high clinical relevance. This should facilitate their implementation, resulting in prescribing optimisation and improved clinical outcomes for multimorbid older adults.