The cross-sectional quasi-experimental study at the University of Szeged, Hungary, worked to determine whether a custom GPT-based chatbot helps first-year international medical students find course information more accurately than a standard text syllabus. The researchers recruited 75 first-year medical students who were international, primarily ESL, and randomly assigned 35 to a chatbot and 40 to a syllabus. Then they were given a 10-minute session followed by a 6-item quiz on the course requirements. The AI was tasked with retrieving information using a custom GPT (GPT-4o) trained on the course syllabus and Q&A. The comparator was a traditional text-based syllabus on the university platform. The median quiz scores for students who used the chatbot were 5.0, and those for students who used the syllabus were 3.5 (P<0.001), indicating that the chatbot users performed better. Supporting this further, 24/35 students who used a chatbot received a score of 5 or higher, while 6/40 of the students who used the syllabus received a score of 5 or higher. The effect size, or Cohen’s d, was equal to 1.29, and the time spent was similar between the two groups. The study had some limitations. Firstly, the ChatGPT prompts couldn’t be analyzed because students used their personal ChatGPT accounts, and the quiz only covered simple course-info questions; the results may not be indicative of deeper learning. Additionally, the study may not generalize to other institutional populations, and it failed to analyze demographic subgroups and fairness.