Background: Negative mental images in social anxiety are often linked to memories of distressing social experiences. Imagery Rescripting (ImRs) has been found to be a promising intervention to target aversive memories, but mechanisms underlying ImRs are largely unknown. The present study aimed (a) to investigate the effects of ImRs compared to cognitive restructuring (CR) on social anxiety symptoms and (b) to extend previous research by examining whether ImRs works by fostering reappraisal of negative emotional self-beliefs. Method: Highly socially anxious individuals (N = 77) were randomly allocated to ImRs, CR, or no intervention control (NIC). A speech task was performed at baseline and at 1-week follow-up. Results: Only CR significantly reduced social anxiety symptoms from baseline to follow-up. Decreases in negative appraisals and emotional distress in response to the speech task did not differ between conditions. Regarding working mechanisms, ImRs led to stronger increases in positive emotions than CR and NIC. Both CR and ImRs yielded short-term reductions in emotionally anchored idiosyncratic self-beliefs, but CR was superior to ImRs at follow-up. Conclusions: The present study provides evidence for the efficacy of a single-session of CR for social anxiety symptoms. As one specific version of ImRs was applied, it is conceivable that other or optimized versions of ImRs might be more effective.