





































 V40 N3 / 2025 

©2025 American Medical Writers Association. All rights reserved.  
ISSN 2163-5315

AMWAJournal.org     4

THEME ARTICLE

ABSTRACT 
Artificial intelligence (AI) is rapidly changing the field of 
health communication. Medical writers, who are central 
to making complex medical information understandable 
and usable, now face both new opportunities and new 
risks. AI can speed up content creation, improve work-
flow efficiency, and scale production. At the same time, it 
introduces concerns related to bias, accuracy, and account-
ability. This paper focuses on 3 core types of bias that affect 
AI-generated content: data-driven bias, algorithmic bias, 
and human bias. These biases often arise from unrepre-
sentative training data, flawed system design, or lack of 
contextual understanding. Left unchecked, they can lead 
to misinformation and worsen health disparities. Medical 
writers play a critical role in mitigating these risks by eval-
uating AI outputs for accuracy, completeness, and fairness. 
When guided by clear standards, collaborative practices, 
and sound editorial judgments, medical writers can help 
ensure that AI supports ethical, equitable, and effective 
health communication. This paper offers practical strategies 
to help medical writers integrate AI tools responsibly with-
out compromising the integrity, ethics, or patient equity of 
health communication.

INTRODUCTION
Health communication shapes how people understand 
medical information, make care decisions, and engage with 
the health system. It influences patients, clinicians, and the 
public, affecting everything from treatment to trust. Medical 
writers are central to this work, translating complex science 
into clear, accurate, and audience-specific content.
 Effective communication requires more than accuracy. 
Language, tone, and context must support understand-
ing and reduce confusion. Poor communication can cause 
harm, whereas strong communication improves outcomes 
and public health.
 Artificial intelligence (AI) is now frequently part of 
health content creation. Large language models (LLMs) 

are used to draft patient materials, clinical summaries, and 
public messages. These tools are fast and consistent, but 
they do not understand science. They rely on patterns, not 
reasoning, and may produce fluent but inaccurate or mis-
leading content.
 As AI becomes more common, concerns about quality 
and accountability grow. Medical writers are often review-
ing or using AI-generated text. They must check for accu-
racy, assess relevance, and intervene when needed. This 
requires a clear grasp of both AI’s strengths and limitations.
 This paper examines common biases in AI-generated 
health content and offers practical strategies for responsi-
ble use. The aim is not to reject AI but to use it in ways that 
uphold the core values of health communication: clarity, 
accuracy, equity, and trust.

BIASES IN AI USE IN HEALTH COMMUNICATION
As AI tools enter health communication, biases become a 
major concern. These biases can arise from training data, 
model design, or how systems are used in practice.1-3 If not 
addressed, they can distort medical information and rein-
force health disparities. To use AI responsibly, we must first 
understand where these biases come from and how they 
affect communication.
 For example, training data often lacks representation 
from diverse populations.3,4 This leads to outputs that ignore 
or misrepresent certain groups, especially those already 
underserved. AI may also “hallucinate” facts or apply 
findings too broadly, further weakening the credibility of 
the content.5,6 Another possible issue is a substantial gap 
between AI developers and health communication experts.3,7 
Without adequate collaboration, AI outputs may fail to meet 
clinical or ethical standards. When human oversight is lim-
ited, flawed content can easily go unnoticed.8 Table 1 pro-
vides detailed information regarding these biases.

STRATEGIES TO MITIGATE AI BIASES
Medical writers play a key role in identifying and correcting 
AI-related bias. Unchecked, these issues can lead to  

Red Thaddeus D. Miguel, MD, MBA, MSc, RAC, RCC; Manal El Joumaa, MSc; Rami Ali, MScPH / Thera-Business, Inc, 
Kanata, Canada

Artificial Intelligence Bias in Health Communication:  
Risks and Strategies for Medical Writers

http://www.amwajournal.org


AMWAJournal.org     5AI Bias in Health Communication

Table 1. Key Biases in AI and Their Implications for Medical Writers
Category Core Issues Examples and/or Evidence Implications for Medical Writers

Data-Driven Bias

Biased Training  
Data Sets

- Nonrepresentative 
training data sets

- Lack of data set 
transparency

- Misclassification or 
omission of demographic 
subgroups

- A scoping review of 70 studies found that AI training 
data sets in dermatology often lacked transparency, 
used unverified labels, and failed to report patient 
diversity.1

- A review of 74 studies on image-based diagnostic  
algorithms revealed a systemic geographic bias, with 
most US models trained on data from just 3 states, 
namely California, Massachusetts, and New York.9

- A scoping review of 7,314 articles published using AI 
techniques found that US and Chinese data sets and 
authors were disproportionately overrepresented, 
amplifying global health inequities by favoring data-
rich regions over data-poor ones.2

- Risk of reinforcing biased or 
incomplete narratives

- Misrepresentation of 
underrepresented populations

- Risk of content lacking global 
applicability

- Risk of amplifying global health 
inequities

Algorithmic Bias

Model Limitations - Design flaws and poor 
generalizability

- Inability to handle 
context or nuance

- Generation of fabricated 
or hallucinated content

- Overfitting the model to a training data set leads to 
poor generalizability to new, unseen cases.4

- AI can confidently generate plausible but fabricated 
data or references.10,11

- A study appraising 2 AI-generated minireviews on 
hereditary angioedema and eosinophilic esopha-
gitis found that although AI used well-articulated 
language, the content lacked depth, analytical insight, 
and contained fabricated references. Despite being 
instructed to use scientific references, the AI chatbot 
relied on freely available resources.12

- Risk of including false or unverifiable 
information

- Inconsistencies in tone, structure, or 
depth

- Compromised credibility of written 
outputs, especially on topics with 
limited resources

Linguistic and 
Cultural Gaps

- English- and Western-
centric training

- Exclusion of non-English–
speaking populations

- Mistranslation 
and cultural 
misrepresentation

- The dominance of English in benchmarks and 
training data for LLMs exacerbates challenges for 
individuals and organizations in the developing 
world, predominantly non-English speakers.13

- Machine translation errors in public-health 
communication included inconsistent use of 
terminology, unidiomatic or awkward style, and 
untranslated text.14

- Miscommunication due to inaccurate 
or culturally inappropriate phrasing

- Barriers to multilingual inclusivity

- Limited accessibility for global 
audiences

Human Bias

Lack of 
Interdisciplinary 
Collaboration

- Limited engagement with 
domain experts

- Limited efforts for data 
exchange

- Unstandardized health 
data systems

- Without standardized terminologies and data 
formats, integrating AI tools into clinical workflows 
becomes challenging.15

- A study assessing health care standards in ophthal-
mology identified multiple gaps, including limited 
adoption of imaging standards, lack of use cases for 
integrating AI-based decision support tools, scarcity 
in common data models to harmonize large data re-
positories, and the absence of standardized interfaces 
and outputs for AI algorithms.16

- The absence of interdisciplinary collaboration in 
AI-driven public-health initiatives leads to a lack 
of standardized classification and summarization 
of both traditional and AI-based methods. This 
fragmentation hampers informed decision-making, 
delays implementation, and deters broader adoption 
of effective tools.7

- Incomplete or contextually flawed 
content

- Lack of alignment with clinical 
realities or public-health messaging

- Inability to critically assess or 
interpret outputs

- Incomplete understanding of AI 
functionality or context

- Undermined reproducibility and 
generalizability of AI-generated 
evidence

Table 1 continued on next page.

http://www.amwajournal.org


AMWAJournal.org     6AI Bias in Health Communication

Human Bias (cont.)

Infrastructure 
Inequity

- Disparities in digital in-
frastructure and funding

- Resource disparities in 
data collection

- Underrepresentation 
of low-income and 
non-Western regions

- Low-resource or underfunded settings are underrep-
resented in data sets and model development due to 
disparities in infrastructure, data access, and resource 
availability.17

- Racial, gender, and age disparities are affecting 
clinical decision-making, quality of treatment, and 
outcome prognosis.18

- Biased messaging rooted in data-rich 
regions

- Exclusion of underrepresented  
communities

- Risk of medical writing failing 
to address the needs of diverse 
populations

Lack of Human 
Oversight

- Overreliance on AI tools 
(automation bias)

- Opaque AI decision-
making (“black box”)

- Inadequate validation 
and ethical safeguards

- Frequent false alarms can desensitize users, leading 
to ignored alerts and perpetuation of AI-generated 
errors (feedback loops).3

- Many AI models function as “black boxes,” making 
their internal logic opaque and limiting humans’ 
ability to interpret outputs.19

- Overreliance on AI tools can diminish human critical 
thinking, leading to uncritical acceptance of AI 
outputs. This issue is exacerbated by human cognitive 
fatigue during sustained oversight tasks.8

- AI systems often depend on large volumes of 
sensitive personal or biological data, raising ethical 
and regulatory concerns. Without effective oversight, 
the balance between data use and individual privacy 
rights cannot be adequately maintained.8

- Overdependence on AI-generated 
text without verification

- Reduced editorial quality and 
integrity

- Risk of ethical or privacy violations  
in published content

AI, artificial intelligence; LLM, large language model.

misinformation and reduce the quality of health communi-
cation. A practical way forward involves combining human 
oversight, collaboration, and clear editorial standards.
 Writers do not need to be programmers to make an 
impact. They can flag biased language, correct errors, and 
ensure that outputs match clinical evidence. They can also 
help ensure that information reflects the needs of diverse 
audiences.
 Working with developers, ethicists, and clinicians 
strengthens the process. Together, these teams can build  
AI tools that are more accurate, inclusive, and context 
aware. Writers can also help create quality control checklists 
and review protocols specific to different health communi-
cation settings.
 Bias cannot be eliminated entirely, but it can be 
reduced. Through careful review and strong editorial judg-
ment, medical writers can guide AI outputs toward accu-
racy, fairness, and relevance.

Human Oversight
Ensuring Accuracy and Clinical Relevance
Despite their remarkable capabilities, AI systems are not 
infallible. They lack human qualities such as cognitive  
reasoning, contextual judgment, and emotional intelli-
gence. These limitations introduce serious risks, especially 
in high-stakes fields like health communication. In this  
context, medical writers serve as critical safeguards respon-

sible for validating the consistency, accuracy, and clinical 
relevance of AI-generated content. This includes checking 
facts, identifying discrepancies, and correcting biased or 
misleading content.
 A common issue when relying on AI-generated data 
is the overgeneralization of specific findings beyond the 
populations studied, especially in summarized medical 
information.20 When AI systems generate diagnostic con-
tent or treatment suggestions, they may generalize find-
ings toward an irrelevant population, without considering 
individual patient needs.21 For example, in a study of chest 
radiograph classifiers trained on 3 large chest x-ray data sets 
and 1 multisource data set, AI systems were found to selec-
tively underdiagnose conditions in underserved patient 
populations.22 The underdiagnosis rates were even higher 
for intersectional subgroups, such as Hispanic female 
patients. These frequent misclassifications increase the risk 
of delayed or missed treatment. This illustrates how algo-
rithmic generalization can perpetuate or amplify disparities 
when AI systems are deployed without accounting for spe-
cific variations or patient context. Medical writers should 
remain alert to this dynamic as overgeneralizations pro-
duced by AI can distort scientific understanding and under-
mine the standards of precision and equity required in 
health communication.
 Another issue is hallucinations, in which even the most 
advanced and well-trained AI tools can generate fabricated 

http://www.amwajournal.org


AMWAJournal.org     7AI Bias in Health Communication

data or references. Hallucination in scientific citation was 
shown to affect approximately 20% to 50% of AI-generated 
content, depending on task complexity.6,23 In response to 
these challenges, medical writers must always scrutinize 
the sources cited and referenced by AI tools. Their over-
sight helps uncover potential inaccuracies or hallucinations, 
which may otherwise go unnoticed.5 This can be facilitated 
by the use of citation verification tools that automatically 
flag fabricated or incorrect references generated by AI.24 
Another effective strategy is applying retrieval-augmented 
generation, which can constrain AI outputs to trusted, real-
time databases, thereby mitigating AI hallucinations.25,26

 Evidence suggests that AI cannot be solely relied upon 
to produce complex health communication materials with-
out human oversight. A recent study by McMinn et al high-
lights the ongoing need for editorial review when using 
LLMs to generate plain language summary abstracts.27 The 
study found that LLMs can introduce persistent errors, such 
as misrepresenting clinical content, reinforcing inaccurate 
associations between conditions and demographics, and 
omitting or misusing sensitive terms. Although AI-assisted 
approaches improved readability and reduced drafting 
time, the authors stressed that medical writers remain 
essential for reviewing and refining content to ensure accu-
racy, clarity, and appropriateness for lay audiences.

Reviewing Fairness, Equality, and Equity
Medical writers must remain vigilant to detect embedded 
biases in AI outputs, such as skewed disease associations, 
exclusion of certain populations, and the use of inequita-
ble language. Although these issues are best addressed at 
the levels of data collection and algorithm design, human 
oversight at the postprocessing stage remains critical. By 
carefully inspecting and revising the generated AI content, 
medical writers can correct language that may exacerbate 
existing inequalities, discriminate against marginalized 
groups, or perpetuate gender stereotypes. Additionally, they 
can mitigate bias through implementing content filters and 
transfer learning methods to adapt models to diverse pop-
ulations. They can also be part of regular audits, continu-
ous monitoring, and feedback loops that are performed to 
enhance fairness, equality, and equity over time.28

 A key element in promoting fairness, equality, and equity 
in health communication is the use of appropriate terminol-
ogy and word choice. AI content does not consistently use 
appropriate or inclusive terminology, which highlights the 
need for supervision by medical writers. For example,  
medical writers can rely on established terminology  
standards such as the International Statistical Classification 
of Diseases, Systematized Nomenclature of Medicine—
Clinical Terms, and Logical Observation Identifiers Names 

and Codes.15 The American Medical Association and 
the Association of American Medical Colleges have fur-
ther emphasized the medical writer’s role in their guide, 
Advancing Health Equity: A Guide to Language, Narrative 
and Concepts, which offers comprehensive support for 
equity-focused, person-first language.29

 Understanding the characteristics of the targeted audi-
ence is essential to effective health communication. However, 
AI may not be able to implement audience-specific writ-
ing in the same way humans do. Medical writers must think 
about who will read their content and how word choice will 
influence interpretation. Accordingly, they must tailor their 
writing tone, structure, and technical depth to suit the target 
audience. The impact of word choice goes beyond clarity; it 
shapes perception and can either foster inclusivity or perpet-
uate exclusion. By being aware of audience needs, both  
medical writers’ expertise and AI can combine to produce 
content that is accurate, culturally sensitive, and tailored to 
diverse audiences.

Training and Education
Several studies suggest that training and educating relevant 
stakeholders can help mitigate AI bias, which can be applied 
to users such as medical writers. Hasanzadeh et al high-
lighted the need for training and educating users on how to 
critically evaluate AI-generated recommendations.30 They 
suggested that routine engagement in critical thinking exer-
cises helps teams recognize and overcome AI pitfalls and 
biases. In addition to improving awareness, these exercises 
help maintain mindfulness of sensitive attributes such as age, 
gender, or ethnicity that may be unintentionally amplified 
in AI outputs. Other studies support this approach, showing 
that training improves how health care professionals use and 
assess AI tools.30-32 One important area for training is explain-
ability, which refers to understanding how AI systems pro-
duce their outputs. Many AI models operate as “black boxes,” 
offering little insight into how decisions are made. In health 
care, in which such outputs can influence care, medical  
writers must be able to assess whether AI-generated content 
aligns with clinical standards. This ability to interpret and 
question AI outputs helps writers ensure ethical, accurate, 
and safe communication.31

Collaboration With Different Stakeholders
Addressing bias in AI requires a coordinated effort across 
health communication stakeholders, including clinicians, 
developers, policymakers, and medical writers.3,17 As digital 
health tools evolve, medical writers must move beyond con-
tent creation and take on more active, collaborative roles.
 From the early stages of AI development, writers can 
help apply clinical terminology appropriately and develop 

http://www.amwajournal.org


AMWAJournal.org     8AI Bias in Health Communication

quality control tools, such as standardized review check-
lists. One example is the model, evaluation, timing, range/
randomization, individual factors, count, and specificity of 
prompts and language checklist, designed to improve con-
sistency in generative AI health care studies by addressing 
model design, evaluation, timing, and other key factors.33 
These tools can be adapted for use in patient education con-
tent, clinical summaries, and public-health messaging.
 Interdisciplinary collaboration also helps align AI out-
puts with current evidence and ethical standards. By work-
ing with ethicists, clinicians, and technologists, writers 
can define limits for appropriate AI use and support policy 
development for transparent disclosure of AI involvement.34

 Writers also play a key role in promoting patient-cen-
tered communication in AI outputs. AI tools may lack 
empathy, but writers can help ensure outputs reflect patient 
needs and values by collaborating with care teams.35 This 
improves engagement, supports individualized care, and 
strengthens trust.
 Finally, medical writers can share best practices through 
conferences and forums. These platforms offer opportuni-
ties to refine how AI is used in health communication, pro-
mote responsible use, and lead training on reviewing and 
editing AI-generated content.

Transparency in AI Usage
Transparency is essential to the ethical use of AI in medi-
cal writing. Writers must take full responsibility for disclos-
ing when AI tools are used, in line with guidelines from the 
International Committee of Medical Journal Editors, which 
state that AI systems do not meet the criteria for author-
ship.36 Authors must also ensure that AI-generated content 
is accurate, free from plagiarism, and properly sourced. 
Because language models may overlook or exclude alterna-
tive viewpoints, writers must actively check for balance and 
ensure content reflects a full range of perspectives.37

 Clear disclosure builds trust. When clinicians, policy-
makers, and the public understand how AI was used to pro-
duce medical content, they are more likely to accept it. This 
includes not only acknowledging AI use but also provid-
ing details about model decisions and data sources. Proper 
citation also gives credit to the creators of the model and its 
training data.38

 Public trust is essential for AI adoption in health care. 
Studies show that lack of transparency leads to skepticism, 
even when the technology offers real benefits.39 Medical 
writers can address this by reviewing AI outputs carefully 
and ensuring that content remains clear, accurate, and 
accountable. Rather than resisting new tools, writers can 
use AI responsibly to strengthen communication and build 
public confidence in its use.

CONCLUSION
As AI continues to shape health communication, the role of 
medical writers must adapt. AI tools can support efficiency 
and help generate various forms of content, but they cannot 
replace human expertise. The risks of misinformation, hal-
lucination, and bias remain significant. Writers must apply 
critical thinking and understand the origins of these risks to 
ensure content remains accurate and responsible.
 AI should be used as a tool, not a source. It can assist 
with early drafting or summarization, but final content must 
be guided by human input. This approach reflects writing by 
design, in which writers take intentional control over how 
content is created, verified, and communicated.
 Keeping a human in the loop is essential. Medical writers 
must stay involved throughout the process to evaluate qual-
ity, correct errors, and ensure the message is clear and equi-
table. Those who learn to work with AI, rather than against 
it, will help lead the field forward. By using AI responsibly 
and thoughtfully, writers can maintain high standards and 
strengthen public trust in digital health communication.

Author declaration and disclosures: The authors note no  
commercial associations that may pose a conflict of interest in 
relation to this article.

Author contact: rmiguel@therabusiness.com

References
1.  Daneshjou R, Smith MP, Sun MD, Rotemberg V, Zou J. Lack 

of transparency and potential bias in artificial intelligence 
data sets and algorithms: a scoping review. JAMA Dermatol. 
2021;157(11):1362-1369. doi:10.1001/jamadermatol.2021.3129

2.   Celi LA, Cellini J, Charpignon ML, et al; for MIT Critical 
Data. Sources of bias in artificial intelligence that perpetuate 
healthcare disparities-a global review. PLOS Digit Health. 
2022;1(3):e0000022. doi:10.1371/journal.pdig.0000022

3.   Cross JL, Choma MA, Onofrey JA. Bias in medical AI: 
implications for clinical decision-making. PLOS Digit Health. 
2024;3(11):e0000651. doi:10.1371/journal.pdig.0000651

4.   Aliferis C, Simon GJ. Overfitting, underfitting and general model 
overconfidence and under-performance pitfalls and best practices 
in machine learning and AI. In: Simon GJ, Aliferis C, eds. Artificial 
Intelligence and Machine Learning in Health Care and Medical 
Sciences: Best Practices and Pitfalls. Springer; 2024. Accessed May 
8, 2025. doi:10.1007/978-3-031-39355-6_10.

5.   Alkaissi H, McFarlane SI. Artificial hallucinations in ChatGPT: 
implications in scientific writing. Cureus. 2023;15(2):e35179. 
doi:10.7759/cureus.35179

6.   Walters WH, Wilder EI. Fabrication and errors in the bibliographic 
citations generated by ChatGPT. Sci Rep. 2023;13(1):14045. 
doi:10.1038/s41598-023-41032-5

7.   Olawade DB, Wada OJ, David-Olawade AC, Kunonga E, Abaire 
O, Ling J. Using artificial intelligence to improve public health: 
a narrative review. Front Public Health. 2023;11:1196397. 
doi:10.3389/fpubh.2023.1196397

8.   Holzinger A, Zatloukal K, Müller H. Is human oversight to AI 
systems still possible? N Biotechnol. 2025;85:59-62. doi:10.1016/j.
nbt.2024.12.003

http://www.amwajournal.org


AMWAJournal.org     9AI Bias in Health Communication

9.   Kaushal A, Altman R, Langlotz C. Geographic distribution 
of US cohorts used to train deep learning algorithms. JAMA. 
2020;324(12):1212-1213. doi:10.1001/jama.2020.12067

10.  Májovský M, Černý M, Kasal M, Komarc M, Netuka D. Artificial 
intelligence can generate fraudulent but authentic-looking 
scientific medical articles: Pandora’s box has been opened. J Med 
Internet Res. 2023;25:e46924. doi:10.2196/46924

11.  Omar M, Nassar S, Hijazi K, Glicksberg BS, Nadkarni GN, Klang E. 
Generating credible referenced medical research: a comparative 
study of OpenAI’s GPT-4 and Google’s Gemini. Comput Biol Med. 
2025;185:109545. doi:10.1016/j.compbiomed.2024.109545

12.  Zybaczynska J, Norris M, Modi S, et al. Artificial intelligence-
generated scientific literature: a critical appraisal. J Allergy Clin 
Immunol Pract. 2024;12(1):106-110. doi:10.1016/j.jaip.2023.10.010

13.  Kshetri N. Linguistic challenges in generative artificial 
intelligence: implications for low-resource languages in the 
developing world. J Global Inform Tech Manag. 2024;27(2):95-99. 
doi:10.1080/1097198X.2024.2341496

14.  Šorak V. From terminological inconsistencies to major 
mistranslations: a qualitative analysis of NMT errors in public 
health communication. AI Linguist. 2024:1(1).

15.  Park HA. Why terminology standards matter for data-driven 
artificial intelligence in healthcare. Ann Lab Med. 2024;44(6):467-
471. doi:10.3343/alm.2024.0105

16.  Baxter SL, Lee AY. Gaps in standards for integrating artificial 
intelligence technologies into ophthalmic practice. Curr 
Opin Ophthalmol. 2021;32(5):431-438. doi:10.1097/
ICU.0000000000000781

17.  Ueda D, Kakinuma T, Fujita S, et al. Fairness of artificial 
intelligence in healthcare: review and recommendations. Jpn J 
Radiol. 2024;42(1):3-15. doi:10.1007/s11604-023-01474-3

18.  Norori N, Hu Q, Aellen FM, Faraci FD, Tzovara A. Addressing bias 
in big data and AI for health care: a call for open science. Patterns 
(N Y). 2021;2(10):100347. doi:10.1016/j.patter.2021.100347

19.  Marey A, Arjmand P, Alerab ADS, et al. Explainability, 
transparency and black box challenges of AI in radiology: impact 
on patient care in cardiovascular radiology. Egypt J Radiol Nucl 
Med. 2024;55:183. doi:10.1186/s43055-024-01356-2

20.  Peters U, Chin-Yee B. Generalization bias in large language 
model summarization of scientific research. R Soc Open Sci. 
2025;12(4):241776. doi:10.1098/rsos.241776

21.  Ullah E, Parwani A, Baig MM, Singh R. Challenges and barriers 
of using large language models (LLM) such as ChatGPT for 
diagnostic medicine with a focus on digital pathology - a recent 
scoping review. Diagn Pathol. 2024;19(1):43. doi:10.1186/s13000-
024-01464-7

22.  Seyyed-Kalantari L, Zhang H, McDermott MBA, Chen IY, 
Ghassemi M. Underdiagnosis bias of artificial intelligence 
algorithms applied to chest radiographs in under-served patient 
populations. Nat Med. 2021;27(12):2176-2182. doi:10.1038/
s41591-021-01595-0

23.  De Wynter A, Wang X, Sokolov A, et al. An evaluation on large 
language model outputs: discourse and memorization. Nat Lang 
Proc J. 2023;4:100024.

24.  Habbal F. The implications and efficacy of online verification 
tools in scientific research and citation practices. Int J Autom Digit 
Transform. 2023;2(1):4-14. doi:10.54878/mnart786

25.  Lewis P, Perez E, Piktus A, et al. Retrieval-augmented generation 
for knowledge-intensive nlp tasks. Adv Neural Inf Process Syst. 
2020;33:9459-9474.

26.  Alkhalaf M, Yu P, Yin M, Deng C. Applying generative AI with 
retrieval augmented generation to summarize and extract key 
clinical information from electronic health records. J Biomed 
Inform. 2024;156:104662. doi:10.1016/j.jbi.2024.104662

27.  McMinn D, Grant T, DeFord-Watts L, et al. Using artificial 
intelligence to expedite and enhance plain language 
summary abstract writing of scientific content. JAMIA Open. 
2025;8(2):ooaf023. doi:10.1093/jamiaopen/ooaf023

28.  Ferrara E. Fairness and bias in artificial intelligence: a brief survey 
of sources, impacts, and mitigation strategies. Sci. 2023;6(1):3. 
doi:10.3390/sci6010003

29.  American Medical Association and Association of American 
Medical Colleges. Advancing Health Equity: Guide on Language, 
Narrative and Concepts. Published 2021. Accessed May 8, 2025. 
https://www.ama-assn.org/system/files/ama-aamc-equity-guide.
pdf

30.  Hasanzadeh F, Josephson CB, Waters G, Adedinsewo D, Azizi Z, 
White JA. Bias recognition and mitigation strategies in artificial 
intelligence healthcare applications. NPJ Digit Med. 2025;8(1):154. 
doi:10.1038/s41746-025-01503-7

31.  Sadeghi Z, Alizadehsani R, Cifci MA, et al. A review of explainable 
artificial intelligence in healthcare. Comput Electr Eng. 
2024;118:109370. doi:10.1016/j.compeleceng.2024.109370

32.  Misra R, Keane PA, Hogg HDJ. How should we train clinicians 
for artificial intelligence in healthcare? Future Healthc J. 
2024;11(3):100162. doi:10.1016/j.fhj.2024.100162

33.  Sallam M, Barakat M, Sallam M. A preliminary checklist (METRICS) 
to standardize the design and reporting of studies on generative 
artificial intelligence-based models in health care education and 
practice: development study involving a literature review. Interact 
J Med Res. 2024;13:e54704. doi:10.2196/54704

34.  Harishbhai Tilala M, Kumar Chenchala P, Choppadandi A, et 
al. Ethical considerations in the use of artificial intelligence and 
machine learning in health care: a comprehensive review. Cureus. 
2024;16(6):e62443. doi:10.7759/cureus.62443

35.  Yelne S, Chaudhary M, Dod K, Sayyad A, Sharma R. Harnessing 
the power of AI: a comprehensive review of its impact and 
challenges in nursing science and healthcare. Cureus. 
2023;15(11):e49252. doi:10.7759/cureus.49252

36.  Defining the role of authors and contributors. International 
Committee of Medical Journal Editors. Accessed May 8, 2025. 
https://www.icmje.org/recommendations/browse/roles-and-
responsibilities/defining-the-role-of-authors-and-contributors.
html

37.  Fingerhut A, Winter DC. Artificial intelligence and medical 
writing: where are we going? Br J Surg. 2023;110(8):883-884. 
doi:10.1093/bjs/znad169

38.  Doyal AS, Sender D, Nanda M, Serrano RA. ChatGPT and 
artificial intelligence in medical writing: concerns and ethical 
considerations. Cureus. 2023;15(8):e43292.

39.  Cheong BC. Transparency and accountability in AI systems: 
safeguarding wellbeing in the age of algorithmic decision-
making. Front Hum Dyn. 2024;6:1421273. doi:10.3389/
fhumd.2024.1421273

http://www.amwajournal.org
https://www.ama-assn.org/system/files/ama-aamc-equity-guide.pdf
https://www.ama-assn.org/system/files/ama-aamc-equity-guide.pdf
https://www.icmje.org/recommendations/browse/roles-and-responsibilities/defining-the-role-of-authors-and-contributors.html
https://www.icmje.org/recommendations/browse/roles-and-responsibilities/defining-the-role-of-authors-and-contributors.html
https://www.icmje.org/recommendations/browse/roles-and-responsibilities/defining-the-role-of-authors-and-contributors.html



