An ASR Confidence Score estimates how certain a speech recognition model is that a recognized word or phrase is correct. Higher scores generally indicate more reliable transcriptions.
Voice AI systems use confidence scores to detect uncertain transcriptions, request clarification, or trigger human review. They help improve reliability in customer-facing voice applications.