We evaluated the advice given by a prototype self-assessment triage system in a university student health centre. Students attending the health centre with a new problem used the automated self-assessment system prior to a face-to-face consultation with the general practitioner (GP). The system's rating of urgency was available to the GP, and following the consultation, the GP recorded their own rating of the urgency of the patient's presentation. Full data were available for 154 of the 207 consultations. Perfect agreement, where both the GP and the self-assessment system selected the same category of advice, occurred in 39% of consultations. The association between the GP assessment and the self-assessment rankings of urgency was low but significant (rho = 0.19, P = 0.016). The self-assessment system tended to be risk averse compared to the GP assessments, with advice for more urgent level of care seeking being recommended in 86 consultations (56%) and less urgent advice in only 8 (5%). This difference in assessment of urgency was significant (P < 0.001). The agreement between self-assessed and GP-assessed urgency was not associated with symptom site or socio-demographic characteristics of the user. Although the self-assessment system was more risk averse than the GPs, which resulted in a high proportion of patients being triaged as needing emergency or immediate care, the self-assessment system successfully identified a proportion of patients who were felt by the GP to have a self-limiting condition that did not need a consultation. In its prototype form, the self-assessment system was not a replacement for clinician assessment and further refinement is necessary.