Download PDF
Semin Neurol
DOI: 10.1055/a-2742-2349
Review Article

AI Prompt Engineering for Neurologists and Trainees

Authors

  • Valdery Moura Junior 

    1   Center for AI and Biomedical Informatics of the Learning Healthcare System (CAIBILS), Mass General Brigham, Somerville, Massachusetts, United States
    2   Department of Medicine, Massachusetts General Hospital, Harvard Medical School, Boston, Massachusetts, United States

  • Peter Hadar 

    3   Department of Neurology, Massachusetts General Hospital, Harvard Medical School, Boston, Massachusetts, United States

  • Shawn Murphy   ‡

    1   Center for AI and Biomedical Informatics of the Learning Healthcare System (CAIBILS), Mass General Brigham, Somerville, Massachusetts, United States
    3   Department of Neurology, Massachusetts General Hospital, Harvard Medical School, Boston, Massachusetts, United States

  • Lidia M.V.R. Moura    ‡

    3   Department of Neurology, Massachusetts General Hospital, Harvard Medical School, Boston, Massachusetts, United States
    4   Department of Epidemiology, Harvard T.H. Chan School of Public Health, Boston, Massachusetts, United States

Funding Information P.H. has received research support from the NIH (UE5/R25) and the American Academy of Neurology/American Epilepsy Society/American Brain Foundation/Epilepsy Foundation (Susan S. Spencer, MD, Clinical Research Training Fellowship in Epilepsy). L.M.V.R.M. has received research support from the NIH (5R01AG073410-02, 2R01AG082693-01), the CDC (5U48DP006377-04-00), and the Epilepsy Foundation (Consultant to the CEO and Director of the Epilepsy Learning Healthcare System).
Further Information
   Permissions and Reprints

Abstract

Large language models (LLMs) have transformative potential in neurology, impacting clinical decision-making, medical training, and research. Prompt engineering, the strategic design of inputs to optimize LLM performance, is essential for neurologists and trainees seeking to effectively integrate these powerful tools into practice. Carefully crafted prompts enable LLMs to summarize complex patient narratives, generate differential diagnoses, and support patient education. In training, structured prompts enhance diagnostic reasoning, board preparation, and interactive case-based learning. Neurological research also benefits, with LLMs aiding in data extraction, computed phenotype generation, and literature synthesis. Despite their promise, challenges remain, including hallucinations, data bias, privacy concerns, and regulatory complexities. This review synthesizes current advances and highlights best practices, including two structured prompt engineering frameworks tailored to neurology: Role-Task-Format (RTF) for routine use and our newly developed BRAIN (Background, Role, Aim, Instructions, Next steps) for complex tasks. We offer practical guidance to maximize accuracy, safety, and equity in LLM outputs, ensuring reliable support for neurologists and trainees.

Keywords

large language models - prompt engineering - neurology - artificial intelligence - clinical decision support

‡ These authors are co-senior authors.




Publication History

Received: 26 July 2025

Accepted: 10 November 2025

Accepted Manuscript online:
11 November 2025

Article published online:
03 December 2025

© 2025. Thieme. All rights reserved.

Thieme Medical Publishers, Inc.
333 Seventh Avenue, 18th Floor, New York, NY 10001, USA