This study compares AI PengTalk’s assessments of Korean children’s pronunciation with the assessments of Korean teachers. Sixty Korean sixth-graders participated as assessees, and four Korean elementary teachers participated as assessors. Both PengTalk and the teachers rated the children’s production of 10 English sentences on a five-point scale. They focused on segmentals, stress-rhythm, intonation, and speech rate. The findings were as follows: Firstly, PengTalk evaluated the children’s pronunciation in the four elements significantly lower than the teachers across all English proficiency levels. Secondly, teachers’ ratings of the students aligned more closely with their pre-evaluated English proficiency levels than the AI PengTalk’s assessments. The teachers rated students at the upper level significantly higher than those at the intermediate level, who were, in turn, assessed significantly higher than those at the lower level in all four elements. Furthermore, AI PengTalk and the teachers differed in the mean order of the four elements, particularly in segmentals. Based on the results of this study, suggestions were made for the development and implementation of AI-based English programs.