Objective: Evaluate how an AI-based virtual triage (VT) and care referral technology impacted live triage and care referral in an outpatient/ambulatory care network.Methods: Analysis of a dataset of 8,088 outpatient online encounters assessed how VT influenced patient care seeking action/behavior.Results: There were modest decreases in patients seeking outpatient care, including in-person or video face-to-face encounters (-12.5%), or engaging self-care (-8.2%). Patient engagement of virtual care through e-visits and telephone calls increased moderately (19.1%). One-third (35.0%) of patients changed care seeking likely as a result of VT care referral. Another third (32.3%) reported a pre-VT care intent aligned with the VT care recommendation, and a third (32.7%) did not change care sought when their pre-VT intent was not aligned. A total of 12.0% de-escalated acuity of care seeking as recommended by VT, most frequently from outpatient care to virtual care (6.5%) or self-care (4.3%). When VT recommended care de-escalation, 53.5% de-escalated care. In 21.2% care acuity was escalated, of whom 10.6% pursued virtual care and 7.5% pursued outpatient care instead of self-care, while 3.1% whose care intent was virtual care instead pursued outpatient care. When VT recommended care escalation, 96.2% escalated care. Overall, 26.7% of patients required no further action or involvement of clinical staff.Conclusions: Virtual triage impacted patient care seeking action/behavior among almost half of patients whose pre-VT intent differed from the VT recommendation, with patients nearly twice as likely to follow recommendations to seek higher rather than lower levels of care acuity, while modestly reducing the number of face-to-face visits and increasing virtual care. Overall, a quarter of patients using VT were able to perform self-care without interacting with the healthcare team. Virtual triage has the potential to efficiently and effectively redirect patients to more appropriate levels of care.