PURPOSE
The widespread application of artificial intelligence (AI) to medical
education has provided new challenges and opportunities for learners and
educators, including in the delivery of anatomy education. Question banks are
an essential learning tool that mirror the assessment style medical students
will encounter throughout their careers, but third-party resources may not
reflect content emphasis and appropriate difficulty for the learner’s
educational stage. We leverage faculty-created course content and the
Gemini-powered AI tool NotebookLM to create a board-style question bank
tailored for first year osteopathic medical students in a Clinical Gross
Anatomy course.
METHODS
In June 2025, lecture notes and slides were uploaded into NotebookLM and a
standard prompt used to generate one question per each session objective for
39 lectures within the Gross Anatomy Course. Faculty content experts reviewed
each question for accuracy, relevance, and clarity. Each question was
accepted, rejected, or lightly modified for inclusion in the question bank.
RESULTS
Of 203 questions generated, faculty experts approved 32% without
modification, edited 46% for clarity or accuracy, and rejected the remaining
22% for inaccuracies, lack of clarity or relevance, and formatting issues,
including pseudocases. The final question banks were released to students,
along with the results of the vetting and editing process and the standard
prompt used to generate questions.
CONCLUSION
Applications of Large Language Models in medical education are powerful
tools to create resources that are content-specific and tailored to the level
of learner. However, faculty oversight is still essential to ensure questions
are high quality, accurate, and meet course expectations. We encourage
transparency when releasing AI-generated content to learners.