Aim Cognitive behavioral therapy (CBT) is recognized as an effective treatment for obsessive-compulsive disorder (OCD). To maximize its effectiveness, we designed an “experimental” CBT defined by the addition of a computerized psychoeducative tool. Method In a participative process involving patients through meetings of the French OCD association (AFTOC) and therapists through methodological workshops, we built a therapeutic tool from an experimental checking task. This task, which had been published in an earlier work, was adapted for its psychoeducative dimension. We here report on a randomized double-blind trial which included 35 patients with a moderate to severe OCD (Yale-Brown obsessive-compulsive scale, YBOCS between 16 and 25) predominant checking symptoms, no comorbidities, and 2-month stabilized or no treatment. Patients were randomly assigned to either “standard” versus “experimental” CBT. Both therapies were conducted by four CBT-experienced therapists specialized in OCD through weekly individualized sessions over 3 months. Therapy sessions of the experimental CBT were conducted as the standard CBT except for a short exercise with the computerized psychoeducative tool performed by the patient and debriefed with the therapist at the end of the sessions. Patients were assessed before, during, after therapy and again 6 months later using standard clinical tools and a neurobehavioral assessment based on an original symptom-provocation task with anxiety ratings including three types of photographs: neutral, generic inducing obsessions (e.g., doorknobs, electric wires…) and personalized (taken by the patients in their own environment). Results Clinically, “standard” and “experimental” CBT resulted in a significant but equivalent improvement (48% vs 45% reduction of the Y-BOCS score; P = 0.36; d = 0.12). Therapists were satisfied with the psychoeducative dimension of the computerized psychoeducative tool but reported variable acceptance across patients. Patients appreciated its usability. The clinical improvement was associated with a reduction of the task-induced anxiety (r = 0.42, P < 0.05), especially towards personalized items (−28,2% vs −20.41% for generic and −6.24% for neutral photographs, P < 0.001). Mid-therapy response level was predictive of the final improvement (r = 0.82, P < 0.001). Conclusion The computerized tool may provide a well-accepted therapeutic adjuvant even though it doesn’t improve the therapeutic effect. Using a personalized symptom-provocation task reveals the parallel evolution of symptoms and neurobehavioral markers through CBT. Despite the difficulty of improving an evidence-based therapy, mid-therapy results call for investigating the possible adjustments of treatment strategies at an early stage.