 |
Using AI to Navigate the Medical
Literature:
Opportunities, Limits, and Best Practices |
Artificial intelligence (AI) tools can help users explore,
summarize, and organize medical literature. However, AI should be used as an
aid to literature discovery and synthesis, not as a replacement for
authoritative databases, critical appraisal, or expert searching.
Key message: Use AI for
efficiency. Use databases for evidence. Use
librarians for
expertise.
AI tools can be useful at several points in the research process,
especially at the beginning of a search when users are clarifying concepts,
identifying terminology, and organizing ideas.
Literature Discovery
- Generate keywords, synonyms, acronyms, and related phrases.
- Suggest broader and narrower concepts.
- Help users translate a clinical or educational question into
searchable concepts.
- Identify emerging terminology that may not be obvious to a novice
searcher.
Example
Searchable Concepts
AI Query: Please provide search terms for the
question Does using AI tutors improve medical student learning?
AI
Results:
| PICO Concept |
Description |
Search Terms |
| Population (P) |
Medical students |
medical student*, medical education, undergraduate medical
education, medical trainee*, health professions student*, Students,
Medical |
| Intervention (I) |
AI tutors and AI assisted educational tools |
artificial intelligence, AI, generative AI, ChatGPT, large
language model*, LLM, chatbot*, AI tutor*, intelligent tutoring system*,
virtual tutor*, adaptive learning, AI assisted learning, AI powered
education |
| Comparison (C) |
Traditional educational methods |
traditional teaching, faculty tutoring, peer tutoring,
conventional instruction, standard curriculum, elearning |
| Outcome (O) |
Learning and educational outcomes |
learning, learning outcome*, academic performance,
examination score*, test score*, knowledge acquisition, knowledge retention,
clinical reasoning, competency, educational outcome*, learner satisfaction,
student engagement, self directed learning |
Search Strategy Support
- Draft preliminary Boolean search strings.
- Suggest controlled vocabulary to verify in MeSH, Emtree, APA
Thesaurus terms, or other database vocabularies.
- Help compare different phrasings of a research question, such as
PICO.
Example
Draft Peliminary Boolean Search Strings.
AI Query: Please provide a PubMed search strategy
for the question How does telemedicine affect diabetes management?
AI
Results:
("telemedicine" OR "telehealth" OR "virtual care")
AND
("diabetes mellitus" OR diabetes)
AND
("glycemic control" OR HbA1c OR outcomes)
Summarization and Organization
- Summarize abstracts or small groups of articles.
- Create draft evidence tables or thematic outlines.
- Convert technical findings into plain language summaries.
Example
Summarize a Small Group of Articles
AI Task: Summarize a small group of articles on
telemedicine and diabetes management.
Example AI assisted summary: Across several
studies, telemedicine interventions were associated with improved access to
care and modest improvements in blood sugar control, although study quality and
patient populations varied.
Example plain language version: Telemedicine may
help some patients manage diabetes more easily, but results differ depending on
the type of program and the patients studied.
Screening Support
AI may assist with title/abstract screening by helping reviewers
prioritize or categorize records. Title/abstract screening is the process of
reviewing article titles and abstracts to decide whether studies appear to meet
predefined inclusion criteria before fulltext review.
Example
Screening Support
Example review question:
"Do telemedicine
interventions improve outcomes for adults with diabetes?"
Inclusion criteria:
- Adults (18 years and older)
- Patients with diabetes mellitus
- Telemedicine or telehealth intervention
- Original research studies
Exclusion criteria:
- Pediatric populations
- Studies not involving telemedicine
- Editorials, commentaries, or reviews
- Non English articles (if specified by the review protocol)
AI may assist by prioritizing citations as likely
include, likely exclude, or needs human review.
| Article Title |
AI Assessment |
Reason |
| Telehealth for Diabetes Self Management in Adults |
Likely Include |
Matches population and intervention criteria. |
| Pediatric Diabetes Education Program |
Likely Exclude |
Study population is pediatric rather than adult. |
| Remote Monitoring in Chronic Disease |
Needs Human Review |
Diabetes and telemedicine use are unclear from the
abstract. |
Important: AI can help
prioritize records and improve efficiency, but final inclusion and exclusion
decisions should always be made by human reviewers.
- Hallucinated or inaccurate citations: AI tools
may invent citations or provide incorrect publication details.
- Incomplete retrieval: General purpose AI tools
do not search the full medical literature.
- Limited transparency and reproducibility: AI
generated answers may change between sessions and may not show exactly which
sources were searched.
- Bias and uneven coverage: AI outputs may reflect
publication bias, language bias, geographic bias, and historical
inequities.
Warning: Never rely on AI
alone to confirm that literature exists.
| Appropriate uses |
Do not use AI alone for |
| Brainstorming keywords and synonyms |
Systematic reviews or scoping reviews |
| Drafting preliminary Boolean logic |
Clinical guidelines or practice
recommendations |
| Identifying controlled vocabulary candidates to
verify |
Meta-analyses |
| Summarizing abstracts after retrieval |
Citation verification |
| Creating plain language explanations |
Determining article quality or risk of bias |
Warning:Never cite an AI
generated reference until it has been verified in PubMed, Web of Science,
Embase, or another authoritative database.
AI may assist with repetitive or exploratory tasks, but humans
remain responsible for judgment, accuracy, interpretation, and
accountability.
- Researchers define the question and eligibility criteria.
- Librarians advise
on database selection, controlled vocabulary, syntax, reproducibility, and
documentation.
- Clinicians and subject experts interpret clinical meaning and
relevance.
- Authors verify citations and take responsibility for the final
product.
| Resource |
Best used for |
How AI can help |
What to verify |
| PubMed customized for
MUSM full text |
Biomedical literature, MeSH, PMIDs, and links to subscribed
content |
Brainstorm keywords, MeSH candidates, and search blocks |
PMID, citation details, MeSH terms, filters, and fulltext
availability |
| Embase |
Biomedical, drug, device, pharmacology, international
literature, and conference abstracts |
Suggest drug synonyms and help translate PubMed concepts |
Emtree terms, conference records, drug/device indexing, and
deduplication issues |
| PsycINFO |
Psychology, psychiatry, behavioral health, social sciences,
and education |
Identify constructs, alternate terminology, and population
descriptors |
APA Thesaurus terms, methodology terms, and subject
coverage |
| Web of
Science |
Citation tracking and interdisciplinary discovery |
Suggest seed papers or topic clusters |
Citation networks, related records, author and institution
details |
| Journal
Citation Reports |
Journal metrics and subject rankings |
Generate questions for evaluating a journal |
Journal Impact Factor, category, quartile, publisher, and
metric context |
| Journalytics |
Journal selection and publishing practices. Identifying
Predatory Journals. |
Suggest possible journal matches to verify |
Journal scope, review process, indexing, fees, and predatory
warnings |
- Accuracy: Verify AI generated facts, citations,
PMIDs, DOIs, journal titles, and publication details.
- Transparency: Record the AI tool used, date
used, prompts or purpose, and review process.
- Privacy: Do not enter protected health
information, confidential research data, unpublished manuscripts, grant
applications, personnel information, or sensitive institutional data into
public AI tools.
- Copyright: Follow publisher, database, and
institutional policies before uploading PDFs, book chapters, or licensed
content.
- Accountability: Human authors and researchers
remain responsible for the final work.
Privacy reminder: For
public AI tools, assume that anything entered may not be appropriate for
confidential, unpublished, patient related, or proprietary information unless
Mercer University/MUSM has approved the tool for that use.
- AI tools can enhance literature searching, summarization, and
discovery, but they should not replace authoritative databases or expert
review.
- Users should verify citations, critically appraise evidence, and
consult librarians
for comprehensive or systematic searches.
- For the "Methods" section of any manuscript, always do the search
in an authoritative database such as PubMed, Web of Science, PsycINFO, or
Embase so that the search is reproducible.
- AI can assist with search strategy, but it is up to the
researcher to create and or modify the search strategy and perform the search
in an authoritative database.
Retrieval Augmented Generation (RAG)
Retrieval Augmented Generation combines a large language model with
an external knowledge source, such as databases, journals, guidelines, or
institutional content. A RAG system retrieves relevant documents and then
generates an answer based on those sources. Examples of Medical AI systems
using RAG or RAG-like approaches include:
ClinicalKey AI,
AMBOSS AI, and Dyna
AI in DynaMedex
Library perspective: RAG systems are strongest when
paired with authoritative sources such as PubMed, Embase, PsycINFO, Web of
Science, and institutional subscriptions.
Explainable AI in Evidence Synthesis
Explainable AI helps users understand how an AI system reached a
conclusion or recommendation. In evidence synthesis, users need to know why
articles were selected, why studies were excluded, and how evidence was ranked
or summarized.
Library perspective: Users should favor AI tools
that provide explanations, citations, and evidence trails.
AI Assisted Systematic Reviews
AI assisted systematic reviews use machine learning and natural
language processing to support search term generation, deduplication,
title/abstract screening, data extraction, and citation management. Human
reviewers remain responsible for final inclusion decisions, data verification,
risk of bias assessment, and interpretation.
Responsible AI and Research Integrity
Responsible AI refers to the ethical, transparent, and accountable
use of artificial intelligence in research and scholarship. Core principles
include accuracy, transparency, privacy, fairness, and accountability.
AI Disclosure Requirements in Scholarly Publishing
Some publishers and journals require authors to disclose whether and
how AI tools were used in manuscript preparation or research workflows. Human
authors are responsible for the accuracy, originality, and integrity of the
final work.
Key Takeaway: Use AI for Efficiency, Databases
for Evidence, and Librarians for
Expertise.
AI can accelerate literature discovery by helping generate search
strategies, identify relevant concepts, and summarize evidence. However,
authoritative databases provide the evidence, and librarians can help provide
the expertise needed to improve search quality, comprehensiveness, and
reproducibility.
Example
Ask AI to Search the Literature
AI Query: Please provide a list of citations, in
AMA style, with PMID numbers, for physician hesitancy to use anesthesia in DCD
organ donation. . Double check to make sure the articles authors, titles,
journals, PMID numbers, etc are totally correct with no mistakes. Provide your
search strategy.
PubMed Search Stategy Provided by AI:
("Donation after
Circulatory Death"[tiab] OR "Donation after Cardiac Death"[tiab] OR DCD[tiab]
OR "non-heart-beating"[tiab]) AND (anesthesia[tiab] OR anesthesiology[tiab] OR
anesthesiologist*[tiab] OR intensivist*[tiab] OR physician*[tiab] OR "critical
care"[tiab]) AND (hesitan*[tiab] OR attitude*[tiab] OR perception*[tiab] OR
concern*[tiab] OR barrier*[tiab] OR ethics[tiab] OR "hasten death"[tiab] OR
analgesia[tiab] OR sedation[tiab] OR palliative[tiab] OR "end-of-life
care"[tiab])
Modify the search strategy as needed and perform the search in
PubMed
Ask AI to Provide a Summary of the Citations
AI Summary:The literature specifically addressing
physician or anesthesiologist hesitancy to use anesthesia during Donation after
Circulatory Death (DCD) is relatively limited. Most publications focus on
ethical concerns surrounding analgesia, sedation, premortem interventions, and
the role of anesthesiologists in DCD rather than outright refusal to provide
anesthesia. Nevertheless, several consistent themes emerge, including concerns
about whether interventions might hasten death, the ethical justification for
procedures performed primarily to facilitate organ recovery, uncertainty
regarding professional responsibilities during DCD, potential conflicts between
patient welfare and organ procurement objectives, variability in institutional
policies, and the importance of preserving public trust in transplantation
programs.
Librarian
Assistance: Librarians can
assist with controlled vocabulary, search design, database selection, and
documentation to improve the quality, comprehensiveness, and reproducibility of
the search.
AI in Medical Literature Searching and Retrieval
- Fiorini N, Canese K, Starchenko G, Kireev E, Kim W, Miller V, et
al. Best Match: new relevance search for PubMed. PLoS Biol.
2018;16(8):e2005343. doi:10.1371/journal.pbio.2005343.
PMID: 30153250.
- Omiye JA, Gui H, Rezaei SJ, Zou J, Daneshjou R. Large language
models in medicine: the potentials and pitfalls: a narrative review. Ann Intern
Med. 2024;177(2):210-220. doi:10.7326/M23-2772. PMID:
38285984.
- Wang L, Wan Z, Ni C, Song Q, Li Y, Clayton E, et al. Applications
and concerns of ChatGPT and other conversational large language models in
health care: systematic review. J Med Internet Res. 2024;26:e22769. doi:10.2196/22769.
PMID: 39509695.
Generative AI in Clinical Practice and Healthcare
- American Medical Association. Augmented intelligence in medicine.
Updated May 20, 2026. Accessed June 16, 2026.
https://www.ama-assn.org/practice-management/digital-health/augmented-intelligence-medicine
- Han Z, Battaglia F, Udaiyar A, Fooks A, Terlecky SR. An
explorative assessment of ChatGPT as an aid in medical education: use it with
caution. Med Teach. 2024;46(5):657-664. doi:10.1080/0142159X.2023.2271159.
PMID: 37862566.
- Maddox TM, Embi P, Gerhart J, Goldsack J, Parikh RB, Sarich TC.
Generative AI in medicine - evaluating progress and challenges. N Engl J Med.
2025;392(24):2479-2483. doi:10.1056/NEJMsb2503956. PMID:
40208922.
- Moulaei K, Yadegari A, Baharestani M, Farzanbakhsh S, Sabet B,
Afrash MR. Generative artificial intelligence in healthcare: a scoping review
on benefits, challenges and applications. Int J Med Inform. 2024;188:105474.
doi:10.1016/j.ijmedinf.2024.105474.
PMID: 38733640.
Retrieval-Augmented Generation (RAG) and Evidence-Grounded AI
- Amugongo LM, Mascheroni P, Brooks S, Doering S, Seidel J.
Retrieval augmented generation for large language models in healthcare: a
systematic review. PLOS Digit Health. 2025;4(6):e0000877. doi:10.1371/journal.pdig.0000877.
PMID: 40498738.
- Gargari OK, Habibi G. Enhancing medical AI with
retrieval-augmented generation: a mini narrative review. Digit Health.
2025;11:20552076251337177. doi:10.1177/20552076251337177.
PMID: 40343063.
- Nanua S, Steward R, Neely B, Datto M, Youens K.
Retrieval-augmented generation for interpreting clinical laboratory regulations
using large language models. J Pathol Inform. 2025;19:100520. doi:10.1016/j.jpi.2025.100520.
PMID: 41244595.
- Perkins G, Anderson NW, Spies NC. Retrieval-augmented generation
salvages poor performance from large language models in answering
microbiology-specific multiple-choice questions. J Clin Microbiol.
2025;63(3):e0162424. doi:10.1128/jcm.01624-24. PMID:
39932275.
- Son N, Kang I, Kim I, Lee K, Nam S, Lee D. Development and
evaluation of a retrieval-augmented generation-based electronic medical record
chatbot system. Healthc Inform Res. 2025;31(3):218-225. doi:10.4258/hir.2025.31.3.218.
PMID: 40840929.
- Zhang S, Phan E, Velmovitsky P, Pham Q, Sanner S.
Retrieval-augmented generation for medical question answering on a heart
failure dataset: performance analysis. JMIR Form Res. 2026;10:e84932. doi:10.2196/84932. PMID:
41747226.
Hallucinations, Citation Accuracy, and Research Integrity
- Resnik DB, Hosseini M. Hallucinated citations produced by
generative artificial intelligence may constitute research misconduct when
citations function as data in scholarly papers. Account Res. Published online
March 15, 2026:2645390. doi:10.1080/08989621.2026.2645390.
PMID: 41833014.
- Zielinski C, Winker MA, Aggarwal R, Ferris LE, Heinemann M,
Lapena JF Jr, et al. Chatbots, generative AI, and scholarly manuscripts: WAME
recommendations on chatbots and generative artificial intelligence in relation
to scholarly publications. Colomb Med (Cali). 2023;54(3):e1015868. doi:10.25100/cm.v54i3.5868. PMID:
38089825.
Bias, Equity, Transparency, and Explainable AI
- Jung J, Lee H, Jung H, Kim H. Essential properties and
explanation effectiveness of explainable artificial intelligence in healthcare:
a systematic review. Heliyon. 2023;9(5):e16110. doi:10.1016/j.heliyon.2023.e16110.
PMID: 37234618.
- Omiye JA, Lester JC, Spichak S, Rotemberg V, Daneshjou R. Large
language models propagate race-based medicine. NPJ Digit Med. 2023;6(1):195.
doi:10.1038/s41746-023-00939-z.
PMID: 37864012.
- Zhang K, Wang D, Lin F, Xie J, Zhou W. A comprehensive review of
explainable artificial intelligence in healthcare methods, evaluation, and
clinical integration. iScience. 2026;29(3):115026. doi:10.1016/j.isci.2026.115026.
PMID: 41858885.
AI-Assisted Systematic Reviews and Evidence Synthesis
- Gargari OK, Mahmoudi MH, Hajisafarali M, Samiee R. Enhancing
title and abstract screening for systematic reviews with GPT-3.5 turbo. BMJ
Evid Based Med. 2024;29(1):69-70. doi:10.1136/bmjebm-2023-112678.
PMID: 37989538.
- Page MJ, McKenzie JE, Bossuyt PM, Boutron I, Hoffmann TC, Mulrow
CD, et al. The PRISMA 2020 statement: an updated guideline for reporting
systematic reviews. BMJ. 2021;372:n71. doi:10.1136/bmj.n71. PMID:
33782057.
Responsible Use, Authorship, and Scholarly Publishing Policies
- Committee on Publication Ethics. Authorship and AI tools.
Published February 13, 2023. Accessed June 16, 2026.
https://publicationethics.org/cope-position-statements/ai-author
- International Committee of Medical Journal Editors. Use of AI by
authors. In: Recommendations for the Conduct, Reporting, Editing, and
Publication of Scholarly Work in Medical Journals. Updated January 2026.
Accessed June 16, 2026.
https://www.icmje.org/recommendations/browse/artificial-intelligence/ai-use-by-authors.html
- International Committee of Medical Journal Editors. Use of
artificial intelligence in publishing. In: Recommendations for the Conduct,
Reporting, Editing, and Publication of Scholarly Work in Medical Journals.
Updated January 2026. Accessed June 16, 2026.
https://www.icmje.org/recommendations/browse/artificial-intelligence/
