The purpose of this article is to explore the application potential of AR (Augmented Reality) technology based on DL (Deep Learning) in the field of education, and to verify its actual impact on students’ learning effect. To achieve this goal, a set of AR education application models with DL algorithm is designed. The model covers many core modules such as data processing, real-time interaction and individualized recommendation. Through experiments in a medium-sized school, the control group was set up to compare the learning effect of students under the traditional teaching method and AR teaching method. The experimental results show that the students in AR education group have obvious advantages in knowledge mastery, problem solving ability and learning efficiency. Compared with traditional teaching methods, the test accuracy of students in AR education group is improved by about 20%, and the response time is shortened by 15%. Furthermore, students’ satisfaction and interest in AR education are generally high, which is helpful to improve their learning enthusiasm and participation. The research shows that the AR education application model based on DL can effectively improve students’ learning effect and learning experience.