Dysfunctional beliefs influence the pathogenesis and maintenance of eating disorders. Cognitive-behavioral therapy (CBT) aims to modify these specific beliefs and mediate functional patterns of appraisal and thinking. The present pilot study aims to experimentally validate a computer-assisted avatar-based treatment (CAT-DB) to reduce dysfunctional body-related cognitions and symptom severity in a sample with subclinical eating disorder symptomatology. In a randomized controlled trial, 48 participants were assigned to either the control group or CAT-DB. The conviction of individual body-related cognitions and eating disorder-specific symptom expression was measured shortly before the beginning of the intervention (pre-measurement), directly after the intervention (post-measurement), and 14 days later (follow-up). There was a significant interaction effect for symptom severity, as reductions in eating disorder-specific symptomatology were significantly greater in the CAT-DB group compared to the control group at follow-up. Compared to the control group, CAT-DB resulted in non-significant but expected improvements in individual body-related automatic thoughts. The present study offers preliminary evidence suggesting that CAT-DB could be efficacious in addressing certain symptoms associated with eating disorders. Further studies with representative clinical samples diagnosed with eating disorders are needed to provide more evidence of the efficacy of CAT-DB as a support for classical CBT in a variety of therapeutic and research contexts.