Aiming at the problems in the experimental teaching of data structure such as the replacement of students’ learning process, the weakening of algorithm thinking, and the distortion of teaching evaluation results caused by the overuse of large models, this paper analyzes the influence mechanism of large models on students’ ability after their intervention in experimental teaching. On the basis of defining the reasonable usage boundary of large models, a new procedural teaching evaluation system is constructed, and an implementation plan for data structure experimental teaching under the constraint of standardized AI usage is proposed. Teaching practice results show that this evaluation model can effectively standardize students’ use of AI and has a significant promoting effect on improving students’ algorithm understanding ability, program debugging ability, and engineering practical ability.