<p>Separation of used paper is a critical issue to increase recovered paper and reduce waste. However, accurate classification of mixed paper remains challenging. The key reason is that there are so many different types of mixed paper. This study proposes an AI-based support system for used paper separation using object detection. The system is developed based on the YOLO (You Only Look Once) framework and is designed to classify 12 categories of recyclable and non-recyclable paper. Experimental results show that the proposed system achieved a mean Average Precision (mAP50) of 0.714 under general conditions. When images without overlapping objects were used, the mAP50 improved to 0.836. These results suggest that the proposed system has high potential to support paper separation tasks. And detection performance was significantly affected by object overlaps. Therefore, when performing object detection on used paper under general conditions where objects may overlap, this factor should be taken into account. As future, we aim to implement this system on a mobile device and promote recycling behavior.</p>