In today’s digital era dominated by e-commerce and technological advancements, traditional offline retail faces significant challenges. The convenience of online shopping platforms has steered consumers away from brick-and-mortar stores, despite their continued economic importance. This study proposes a shopping recommendation system that utilises the ResNet-50 neural network for face recognition, combined with the Apriori algorithm, to segment customer data by age and gender. By mining customer transaction records, the system organises and arranges items based on the extracted information, and strategically divides store areas. It then recommends customers to the corresponding zones, thereby saving the customer time, enhancing the shopping experience, and boosting sales. The MAE of face recognition age is 5.38, the accuracy of gender recognition is about 85%, and the average response time of the system per frame is 0.4ms. This research aims to rejuvenate offline retail through strategic AI integration, fostering unique customer interactions and operational efficiency.