Cognitive behaviour therapy (CBT) is now the psychological treatment of choice for psychosis but meta-analyses indicate a low effect size on delusions, so further innovations are clearly needed, and group CBT for psychosis (GCBTp) is an under-researched area. This study aimed to service-evaluate the feasibility, satisfaction, safety, and effectiveness of a CBT group specifically targeting medication-resistant single delusions in early psychosis patients (EI-GCBTp). Three separate EI-GCBTp groups were run resulting in a total of 11 medication-resistant early psychosis patients. A within-subjects design tested for group change across two time points: pre-baseline (4 weeks before treatment) to baseline (session 1 of treatment) and sessions 1–8 (the treatment period). Thirteen delusion dimensions were measured from three psychosis-specific questionnaires: The Psychotic Symptom Rating Scale (PSYRATS), Characteristics of Delusion Rating Scale, and the Belief Rating Scale. At least three patients attended each group, satisfaction scores were high, and no harm to patients was identified. With reference to effectiveness, the pre-baseline period showed virtually no change. In contrast, across the EI-GCBTp treatment period, the PSYRATS total demonstrated a statistically significant decrease in delusional severity (p < 0.01), a 31% symptom reduction, and a large effect size (Cohen's d = 1.2, 95% confidence interval = −2.53 to 0.05), statistically significant across four delusion dimensions. EI-GCBTp appears feasible, acceptable, safe, and preliminary uncontrolled effectiveness results suggest merit for larger-scale more rigorous testing of this treatment format for possible dimensional improvements of persistent delusions.