Objective It is necessary for suicide prevention interventions to demonstrate safety. One important aspect of safety is evidence that deterioration rates are low. No studies have examined deterioration of suicidal ideation in the context of an internet-based cognitive behavioral therapy (CBT) intervention to reduce suicidal ideation. A secondary analysis of the outcomes of an online trial was conducted to determine rates of deterioration in a CBT intervention called Living with Deadly Thoughts, relative to an attention control condition, and to identify factors associated with deterioration. Method The randomized controlled trial included 418 adults with suicidal ideation at baseline (77% females, mean age 40.6 years). Deterioration was defined in two ways: a reliable increase in Suicidal Ideation Attributes Scale (SIDAS score) of ≥6.0 units; or triggering the safety protocol during the trial as determined by high levels of suicidality. Analyses were repeated with multiply imputed data. Predictors of deterioration were identified using logistic regression analysis. Results 30 (14%) participants in the control group and 15 (7%) in the intervention group met criteria for reliable deterioration (Fisher’s exact p = 0.027). In a fully adjusted logistic regression model, greater initial severity of suicidal thinking was associated with lower odds of deterioration, while intervention condition, demographics, psychological factors and mental health factors had no significant association with deterioration. Conclusions Participation in an online suicide prevention intervention was associated with lower prevalence of reliable deterioration than participation in an attention-control intervention, providing further evidence that internet-based CBT interventions do not cause harm.