Schizophrenia is a serious mental illness for which the mainstay of treatment is antipsychotics. Up to 30% of schizophrenia patients show limited response to antipsychotics. Identifying these patients before treatment could guide individualized treatment for improving outcomes in those not likely to show robust benefit from antipsychotics. Diffusion tensor imaging was performed with 56 drug-naïve first-episode schizophrenia patients and 69 matched healthy controls. Patients were followed clinically after one-year of antipsychotic treatment and classified at that point into groups of 17 poor outcome and 39 good outcome patients based on whether they showed at least a 50% reduction of Positive and Negative Syndrome Scale (PANSS) scores from baseline. Tract-based spatial statistics were applied to assess white matter microstructure in the two patient subgroups and healthy controls. Poor outcome patients showed reduced pretreatment fractional anisotropy (FA) in left cingulum and anterior thalamic radiation and increased FA in right superior and inferior longitudinal fasciculus compared with good outcome patients. FA in each of these four tracts was decreased in both patient subgroups relative to healthy controls. Considered together, the four altered tracts showed promising ability to differentiate poor from good outcome patients (sensitivity = 74.4%, specificity = 95.2%, AUC = 0.90, p < 0.001), and superior prediction of clinical outcome to baseline PANSS scores (p < 0.015). Prediction of outcomes using DTI features was not related to duration of untreated psychosis. Baseline alterations in white matter integrity may identify schizophrenia patients less likely to respond to treatment, which could be useful information for stratification in clinical trials and for individualized treatment planning.