OBJECTIVE: Demographic, behavioral, and diagnostic information should routinely be collected from clients with severe mental illness, and data gathering should employ the most efficient techniques available. Surveys are increasingly conducted via Web-based computer-assisted interviewing (CAI), but this technique is not well validated for patients with severe mental illness. A randomized clinical trial of 245 clients was carried out to compare face-to face and computer-assisted interviewing (233 clients completed two surveys). METHODS: Self-report data were collected on demographic characteristics, substance abuse, risk behaviors for blood-borne diseases, trauma history, and posttraumatic stress disorder. Each client was assessed twice and randomly assigned to one of the four possible combinations of interviewer and computer (computer and computer, N=53; computer and interviewer, N=56; interviewer and computer, N=59; and interviewer and interviewer, N=65). The two formats were compared on feasibility, client preference, cost, reliability, convergent validity, and criterion validity. RESULTS: This study demonstrated the feasibility of CAI across a variety of inpatient and outpatient settings. All participants who began the CAI process completed the interview and responded to over 95% of the survey items. Participants liked using the computers as well as they liked face-to-face interviews, and they completed the CAI as quickly. CAI produced data as reliable and valid as face-to-face interviews produced and was less expensive, and results were available more quickly. The two formats were similar in criterion validity. CONCLUSIONS: CAI appears to be a viable technology for gathering clinical data from the population with severe mental illness and for transforming such information into a useful, quickly accessible form to aid in clinical decision making.