The blended learning approach has been explored many times in theory and seems to be argued many times about its superiority, but the effect in practice still lacks effective data to prove it. This paper takes full account of the contingencies that exist in teaching practice, focuses on English-speaking teaching as the object of study, and improves the teaching control design. Integrate the concept of blended teaching and clarify the overall mechanism of blended teaching in spoken English. Ensure that the experimental group and the control group have comparable levels of spoken English through random selection, and try our best to exclude the interference of other factors in the teaching experiment. On the other hand, we build an objective model of teaching quality by combining the BP neural network and genetic algorithm. The teaching effect was fully verified by integrating students’ performance and quality evaluation results. The results show that before and after the blended teaching, the students in the control group and the experimental group have different degrees of improvement in satisfaction, learning motivation, and attitude. The model’s evaluation of teaching quality before and after the teaching also validates the effectiveness of the blended teaching mode in teaching spoken English.