This study assesses how effectively an AI training module improves pediatric fracture identification among medical trainees. A total of 240 pediatric upper-extremity radiographs were divided into two balanced groups based on anatomy, fracture presence, and displacement. 4 medical students and 4 radiology residents were recruited to interpret the images. The AI-based training module utilized a previously validated open-source fracture detection algorithm. Participants were tested on one of the two groups (A or B) and asked to indicate whether they thought a fracture was present, its location, and their confidence level from 0-100. Accuracy, sensitivity, and specificity for fracture detection were also calculated for each reader group. After the initial testing, participants reviewed AI-generated annotations for the first set of cases. They then completed the second set without AI assistance to evaluate how the training module influenced their diagnostic accuracy. Findings showed that the radiology residents demonstrated a significant improvement in diagnostic accuracy (71.3% to 77.5%), particularly in identifying non-angulated and non-displaced fractures. In contrast, medical students did not improve substantially (56.3% to 57.3%). Confidence scores for both groups remained largely unchanged (residents 79.0% to 80.6 and students 46.1% to 46.3%). Overall, the findings indicate that the effectiveness of AI-based training depends on the trainee’s prior knowledge and experience in the field.