Objective: To explore pretreatment and short-term improvement variables as potential moderators and predictors of 12-month follow-up outcome of unsupported online computerized cognitive behavioral therapy (CCBT), usual care, and CCBT combined with usual care for depression. Method: Three hundred and three depressed patients were randomly allocated to (a) unsupported online CCBT, (b) treatment as usual (TAU), or (c) CCBT and TAU combined (CCBT&TAU). Potential predictors and moderators were demographic, clinical, cognitive, and short-term improvement variables. Outcomes were the Beck Depression Inventory-II score at 12 months of follow-up and reliable change. Results: Those with higher levels of extreme (positive) responding had a better outcome in CCBT compared with TAU, whereas those having a parental psychiatric history or a major depressive disorder diagnosis had a better outcome in CCBT&TAU compared with TAU. Predictors regardless of treatment type included current employment, low pretreatment illness severity, and short-term improvement on clinical variables. Conclusions: Optimistic patients, holding approach-oriented coping strategies, might benefit most from CCBT, whereas CCBT&TAU might be the most suitable option for those with more severe vulnerability characteristics. Those with the least impairment improve the most, regardless of treatment type. (Contains 3 tables and 1 figure.)