With the application of artificial intelligence or AI technology in electric power inspection, the defect detection rate of transmission inspection image analysis models has reached 85%–90%. The AI analysis model almost covers all defect classes of transmission lines. But there are a large number of false detection boxes, making it difficult to replace manual defect detection in images. It requires workers to examine boxes predicted by AI models to correct missed detection boxes and false detection boxes. This article proposes a defect review method after artificial intelligence model analysis of electric power inspection images. We optimize the review process by utilizing the similarity of missed detection boxes and false detection boxes on the same transmission line. It can reduce working time, and improve defect detection rate. By using this method, workers can examine all predicted boxes with confidence probability bigger than 0.1 by using only one-third of traditional working time, and increase the defect detection rate to 95%.