Although Internet-based cognitive behavior therapy (iCBT) is an effective treatment for chronic tinnitus, several patients do not improve. In the current study, baseline and process variables were compared between non-responders and responders. Data from patients participating in two randomized controlled trials on iCBT for chronic tinnitus were re-analyzed. Based on the literature, a pre-post difference on the Tinnitus Handicap Inventory (THI) of less than seven points improvement was used to operationalize non-response. Associations between non-response and baseline variables (age, gender, and questionnaire scores), patient progress (THI), the process of the therapeutic alliance (Working Alliance Inventory-Short Revised; WAI-SR), as well as other process variables (number of logins, amount of messages sent from therapists to patients) were investigated. The results showed that non-responders had a less favorable change on the THI than responders already at mid-treatment (pamp;lt;.05). The alliance (WAI-SR) during iCBT was not associated with non-response. Non-responders showed more severe sleep disturbances, logged in less in the iCBT platform, and received fewer messages from the therapists than responders, but these differences were mostly not significant anymore when correcting for multiple testing. To conclude, no symptom change in the first half of iCBT for chronic tinnitus patients is a risk factor of not benefiting from iCBT.