2025 American Academy of Neurology Abstract Website

Objective:

Evaluate how effective an AI-powered research assistant (NotebookLM) is in generating a “Podcast-like” review of landmark articles in Neurology.

Background:

The Artificial Intelligence (AI) Era is upon us and Medical Educators are wondering how to incorporate AI into graduate medical education. Much of the attention has been focused on text generation and less so on audio generation.

Design/Methods:

We selected 5 landmark articles in Neurology published in the last 8 years. A PDF of each article was uploaded into NotebookLM. The program generated a conversation based on the article provided. Two “conversations” were drafted for each article. Both conversations were reviewed by a Neurologist who chose the higher quality conversation for inclusion in the study. Three Medical Students were then tasked to listen to the conversations and grade them accordingly.

Results:

The NotebookLM generated article reviews that were selected were between 7-10 minutes long. All conversations were between a male and female voice. Responders felt that the ideal audience for the conversations was medical students (73.3% ), followed by laypeople (26.7%). Responders were very satisfied with the audio quality (73.3%). Overall strengths of the conversations included content clarity and the efficiency of reviewed content. Areas for improvement included the depth of analysis and length of conversations. Overall, responders felt that the conversations generated were the of same quality or somewhat better than traditional podcasts (90%). Lastly, responders would recommend these conversations to a colleague 73.3% of the time.

Conclusions:

This program highlights the potential future of AI-generated podcasting which can be taken advantage of for trainees who are more audio-based learners. Content clarity is already of high quality, but there is still room for improvement regarding more detailed and comprehensive conversations with future versions of NotebookLM.