Skip to main content
NCBI home page
Wiley Open Access Collection logoLink to Wiley Open Access Collection
. 2026 Mar 10;58(2):e70076. doi: 10.1111/jnu.70076

Development and Evaluation of a Generative AI Chatbot for Database Searching in Systematic Review

Wai San Wilson Tam 1, Neo Tung 2, Shi Xuan Lee 1, Gregor Štiglic 3, Tom Huynh 4, Arthur Tang 4,✉
PMCID: PMC12976166  PMID: 41808290

ABSTRACT

Introduction

Systematic reviews (SRs) require comprehensive, reproducible searches, yet developing search strategies is resource‐intensive and demands specialized expertise. Generative AI offers potential to streamline this process, but empirical evaluations for GAI‐assisted SR searching remain scarce. The objectives of this study are to: demonstrate a step‐by‐step process for developing a custom ChatGPT‐based chatbot to support SR search strategy development, and evaluate its performance.

Design

A cross‐sectional evaluation study.

Methods

We used ChatGPT‐4.0 to create a chatbot designed to mimic a medical librarian, generating PICO‐informed searches. Its knowledge base was augmented with two methodological references. After piloting testing, we refined its instructions. For evaluation, we randomly sampled 50 Cochrane SRs published in 2024. Standardized P–I–O prompts produced database‐ready queries for PUBMED and EMBASE. The primary outcome was per‐review success rate, summarized by median and inter‐quartile range. A sensitivity analysis was conducted.

Results

Pilot testing achieved a retrieval rate of 41/49 (83.7%). In the main sample (1169 studies; median 13.5 studies per SR), the chatbot identified a median of 67.4% of included studies (IQR: 43.1%–88.4%). When limited to indexed studies (n = 1114), retrieval rose to 72.0% (IQR: 46.0%–92.5%). Lower performance was observed when outcomes were absent from the abstracts or interventions had many lexical variants.

Conclusions

A GAI‐based chatbot can rapidly generate SR searches (~67%–72% identification), serving as a useful starting point but not a replacement for expert‐led approaches. Integration of librarian expertise, structured prompts, and controlled vocabularies may improve performance. Further benchmarking and transparent reporting are needed to guide adoption.

Keywords: database searching, generative artificial intelligence, large language model, systematic review

1. Introduction

Systematic reviews (SRs) are essential for evidence‐based healthcare, providing a rigorous synthesis of all available research on a focused question (Glasziou et al. 2001; Sutton et al. 2009). Their impact is especially profound in nursing, where high‐quality evidence is crucial for improving patient outcomes and optimizing care delivery. However, the validity of SRs rests immensely on the comprehensiveness of their search strategies. Comprehensive searches often require over 100 h of skilled work by trained information specialists (Erwin 2004; Saleh et al. 2014), and as the volume of biomedical literature grows, this challenge has intensified. Although structured frameworks, such as the 15‐step search method developed by Bramer et al. (2018), have improved the rigor of search design, the sheer scale and complexity of modern SRs make the process a major bottleneck, particularly for nursing researchers.

Recent advances in generative artificial intelligence (GAI), such as ChatGPT, have opened new opportunities to reimagine how SR searches can be developed. GAI models can generate natural language outputs, dynamically craft search strings, and even simulate reasoning. Early work has explored their use in nursing education and nursing research support (Tam et al. 2023; Woo et al. 2024; Kwok et al. 2024; Tran et al. 2024; Wei et al. 2024; Bui et al. 2025; Tang et al. 2023; Gosak et al. 2025). Emerging GAI platforms now allow users to create custom chatbots tailored to specific research workflows. Such tools could assist with multiple stages of the SR process, including database searching, study selection, and risk of bias assessment, offering a way to augment human expertise.

Despite this potential, practical guidance and real‐world evaluations remain scarce. Few studies have systematically tested how GAI performs when tasked with generating search strategies, but there is little evidence on its reliability compared to expert‐led approaches. To address this gap, we conducted a study with two goals:

  1. to demonstrate a step‐by‐step process for building a custom ChatGPT‐based chatbot to support SR search strategy development, and

  2. to evaluate its retrieval performance using a sample of 50 Cochrane intervention SRs published in 2024. By doing so, we aim to provide early evidence of the feasibility and potential value of GAI‐driven tools in supporting the search component of SRs in nursing research.

1.1. Design

A cross‐sectional evaluation design was adopted to evaluate the performance of the chatbot.

2. Methods

2.1. Chatbot Development and Initial Settings

We created a custom chatbot using OpenAI's ChatGPT‐4.0, following the procedures described by Inita (2024). The chatbot was designed to simulate the role of a medical librarian and generate comprehensive search strategies for SRs. The initial set of instructions provided to the chatbot were as follows:

  1. Role definition: Assume the role of a medical librarian with expertise in searching various academic databases.

  2. Search construction: Utilize the Population (P), Intervention (I), Control or Comparator (C), and Outcome (O) provided by users to create a comprehensive search strategy in databases such as PUBMED, EMBASE, and CINAHL.

We applied a role‐playing prompt engineering approach, tasking the chatbot with conducting a search based on a published systematic review from the Cochrane Database of Systematic Reviews (Ngai et al. 2016). The Population, Intervention, and Outcomes from this SR were used to generate the following initial test prompt:

Create me a search string based on: Population is COPD patients, Intervention is Tai Chi exercise, and Outcome is physical functions or psychological outcomes. For comprehensive search, use both subject headings and text words (keywords). Also build me a complete search string for PUBMED.

The initial results were suboptimal, so we enhanced the chatbot's knowledge base by uploading two methodological references on SR search design (Bramer et al. 2018; Lefebvre et al. 2008). This iterative refinement led to improved outputs, which were judged reasonable for pilot evaluation. Additionally, we added a third instruction to refine its approach:

  • iii

    Search sensitivity: Try to provide a more comprehensive search; prefer to be more sensitive instead of specific in the search.

Following these enhancements, we re‐entered the same test prompt. The resulting output demonstrated marked improvements in structure and coverage, supporting its readiness for subsequent evaluation. Figure 1 shows how the prompt works in the Chatbot and the developed search strategy can be directly copied to PUBMED.

FIGURE 1.

FIGURE 1

Open in a new tab

The prompt and result from the Chatbot.

2.2. Pilot Evaluation and Refinement

We piloted the refined Chatbot three published Cochrane Reviews for interventions (Choi et al. 2010; Cheng et al. 2017; Ngai et al. 2016) to assess whether the search strategies could identify the included studies in PUBMED and EMBASE. We limited the comparison to these two databases as they are among the most commonly used databases in SRs. To ensure consistency, we limited the comparison to English‐language articles, excluding dissertations, trial registries, research reports, and data files.

The initial searches identified 41 of 49 included studies (83.7%), indicating promising retrieval rates. A medical librarian then reviewed the strategies and provided additional guidance to further improve comprehensiveness:

  • iv

    Use of subject headings and keywords: For comprehensive search, use both subject headings and text words (i.e., searching for keywords in Title/Abstract fields).

The workflow for the chatbot is illustrated in Figure 2, and the final optimized chatbot, NUS Nursing Database Searching Chatbot, is publicly available at: https://chatgpt.com/g/g‐67c6e70530c0819189c6825c2bc4b008‐nus‐database‐searching‐chatbot.

FIGURE 2.

FIGURE 2

Open in a new tab

Workflow of the NUS Nursing Database Searching Chatbot. Users provide key PICO elements (Population, Intervention, and Outcome), which are processed by a custom ChatGPT‐4.0–powered chatbot configured to simulate a medical librarian and generate systematic review (SR) database search strategies. The chatbot outputs tailored search strings for multiple databases, including PubMed, Embase, and CINAHL.

2.3. Evaluation

We conducted a formal evaluation of the GAI–assisted chatbot designed to develop search strategies for SRs of interventions. The assessment involved a random sample of Cochrane SRs published in 2024 and followed a standardized, PICO‐informed prompting and retrieval procedure across two bibliographic databases.

2.4. Data Sources and Eligibility Criteria

  • Source of SRs: Cochrane Database of Systematic Reviews.

  • Search platform: PubMed.

  • Timeframe: SRs published in 2024.

  • Eligible SR types: Completed SRs of interventions.

  • Exclusions: Protocols, scoping reviews, SRs focusing on observational designs or diagnostic test accuracy

2.5. Sampling Procedure

A PubMed journal search retrieved 354 records published in the Cochrane Database of Systematic Reviews in 2024. After removing ineligible record types (one scoping review, one protocol), each remaining SR was assigned a random number between 1 and 1000 using Excel's RANDBETWEEN() function. The 50 SRs with the smallest random numbers were selected for evaluation (Figure 3).

FIGURE 3.

FIGURE 3

Open in a new tab

Sampling procedure for the study.

2.6. Search Strategy Generation and Execution

For each selected SR, we extracted the Population, Intervention, and Outcome(s) (P–I–O) elements from the review question and inclusion criteria. Using these elements, we constructed a structured prompt and tasked the chatbot with generating database‐ready search strategies for PUBMED and EMBASE in the following format:

Create me a search string based on: Population is __________, Intervention is __________, and Outcome is ____________. For comprehensive search, use both subject headings and text words (keywords). Also build me a complete search string for PUBMED.

Please note that users are only required to specify the Population, Intervention and Outcomes based on the PICO developed from the original research questions.

Once the search strategy was developed by the chatbot, it was copied to the database (e.g., PUBMED) for performing the search. Full prompt templates for all the Cochrane reviews are provided in the Data S1.

2.7. Reference Identification and Matching

We compiled a list of included studies which are reported in each SR and applied the following criteria:

  • Published in English; and

  • Reported as journal articles, excluding trial registry entries, theses/dissertations, and research reports.

We then determined whether each included study was identified by the AI Chatbot‐generated search results from either PubMed or Embase.

For the evaluation, we copied the AI Chatbot generated search strategy to the database, say PUBMED, and performed the search to obtain the search results. We then identified each included article in the SR by title search. Once the article is found, we used the “AND” operator in PUBMED to combine the search results and the article to see if the search results covered that article. We repeated this step for all the included studies in the SR. Once we completed all the included articles in the SR, we computed the retrieval percentage for that SR. A study was considered “identified” if it appeared in the results from at least one of the two databases.

2.8. Outcome Measures and Analysis

  • Primary outcome: Success rate per SR, defined as the proportion of included English‐language journal articles identified by the searches divided by the total number of included English‐language journal articles in that SR. For example, if there were 12 included studies in the SR, we check one by one how many of them were in the search results, say 10; then, we compute success rate as 10/12 = 83.3%.

  • Summary statistics: The success rates were summarized across the 50 SRs using the median and interquartile range (IQR).

3. Results

We evaluated a total of 50 Cochrane SRs published in 2024, the list of the SRs is shown in Data S1. Across these SRs, 1169 eligible studies (English‐language journal articles) were identified for the assessment. The number of eligible studies per SR varied considerably, with a median of 13.5 studies (IQR 6.0–38.5).

The median success rate for identifying eligible included studies using chatbot‐generated searches was 69.2% (IQR 45.3%–88.4%) across the 50 SRs (see Table 1). Sensitivity analysis was conducted, restricting the denominator to studies indexed in PUBMED or EMBASE (n = 1114). Under this condition, the median success rate increased to 72.1% (IQR 47.6%–92.5%). We further divided the SRs related to Pharmacological/Clinical and Non‐Pharmacological where the median success rate was 78.6% and 72.1% respectively (p = 0.939).

TABLE 1.

Summary results for the NUS Nursing Database Searching Chatbot.

Median (IQR) Median success rate (IQR)
Number of articles identified in PUBMED search per systematic review 974.5 (123.5–4357.0) —
Number of articles identified in EMBASE search per systematic review 2113.0 (482.0–13199.0) —
Number of eligible articles per systematic review 13.5 (6.0–38.5) 69.2% (43.8%–88.9%)
Number of eligible articles per systematic review (excluding those not in PUBMED or EMBASE) 13.0 (3.0–33.0) 72.1% (47.6%–92.5%)
Subgroup
Pharmacological or Clinical Intervention (n = 31 SRs) 78.6% (46.7%–93.3%)
Non‐pharmacological (n = 19 SRs) 72.1% (50.1%–84.6%)
Open in a new tab

We examined cases with notably low retrieval performance to better understand limitations. For instance, one systematic review (Trivedi et al. 2024) included a single primary study (Rao et al. 2022), but the primary and secondary outcomes reported in the review were not mentioned in the primary study's abstract, making retrieval difficult. Unless the search strategy was limited to population and intervention terms alone, the study could not be identified, resulting in a 0% success rate for this review. In another case, a network meta‐analysis (Lax et al. 2024) on topical anti‐inflammatory treatments for eczema had a success rate of only 39%. This lower performance was largely attributable to the numerous variations in anti‐inflammatory treatment terms, which complicated comprehensive search coverage.

4. Discussion

SR aims to synthesize all available evidence from the academic literature to answer a focused research question. A comprehensive and reproducible search strategy is the foundation of a high‐quality SR. Achieving both sensitivity in capturing all relevant studies and precision in minimizing irrelevant results can be challenging. Yet, developing such strategies remains a highly specialized task, often requiring significant time and expert input. Traditional approaches, such as manually compiling synonyms, adapting search strings from published SRs, and seeking librarian consultation, remain valuable but are resource‐intensive and can introduce variability.

4.1. Interpretation of Findings

This study explored whether GAI could help bridge this gap by assisting in a part of the search strategy development process. By creating a custom ChatGPT‐based chatbot and evaluating its retrieval performance on 50 Cochrane reviews, we demonstrated that the chatbot could identify approximately two‐thirds to three‐quarters of relevant studies. While this performance falls short of what would be expected from a carefully curated, librarian‐developed strategy, it provides a pragmatic starting point for researchers, particularly those working without immediate access to specialized expertise such as medical librarians.

The 72% retrieval success rate among indexed studies highlights both the potentials and the current limitations of GAI. Performance shortfalls in some cases were primarily attributable to two factors. First, some studies were not retrieved because key outcomes were absent from abstracts or inadequately indexed, limiting discoverability without manual adjustment. Second, topics with highly variable or inconsistent terminology, such as the eczema network meta‐analysis, challenged the model's ability to comprehensively capture all relevant synonyms. These findings suggest that, although GAI can rapidly generate viable search strategies, expert human oversight remains essential, particularly for complex or terminologically heterogeneous review questions. Beyond retrieval performance, time efficiency represents a substantial practical advantage. Bullers et al. (2018) reported a median of 5.0 h to develop a search strategy among 105 librarians with experience in at least one systematic review. In contrast, a chatbot can generate an initial strategy within minutes when the PICO framework is specified, a reasonable assumption, as PICO is typically provided to librarians during systematic review workflows.

4.2. Implications for Nursing Research

Our results align with broader research on integrating GAI into SR workflows. Similar to work on AI‐assisted critical appraisal (Stiglic et al. 2025), GAI can streamline discrete tasks but cannot fully replace expert judgment. Importantly, structured prompt design appears to improve output quality. Approaches like the CO‐STAR framework (GovTech 2023) may help researchers craft more targeted prompts, further enhancing search precision and relevance. As underlying models continue to evolve, performance is also expected to improve. This study was conducted using ChatGPT 4.0, and future models are likely to offer greater reliability, precision, and sophistication in supporting systematic review workflows.

For nursing and healthcare teams with limited resources, a GAI‐powered chatbot could assist and accelerate the early stages of SR development, freeing time for other tasks such as screening and synthesis. However, these tools should be viewed as augmentative rather than substitutive. Integrating librarian review and domain‐specific ontologies (e.g., MeSH, Emtree) remains critical to ensure completeness and reproducibility. Furthermore, nursing SRs often span interdisciplinary interventions, increasing the complexity of search strategies and the risk of missed studies if AI outputs are not carefully validated.

The ethical dimensions of using AI in scientific research have been heavily debated in recent years, with Resnik and Hosseini (2025) offering a helpful synthesis of key concerns relevant to our work. First, reproducibility is fundamental to rigorous inquiry, yet chatbot‐generated search strategies can vary across sessions, undermining replication. Second, researchers may be held responsible for misconduct stemming from reckless reliance on AI; accordingly, any AI‐generated search strategy should be independently checked and validated before implementation. Third, the “black box” nature of AI models—driven by vast datasets and complex parameters—poses significant challenges for transparency and trust. A core mitigation is clear disclosure of AI use (Tang et al. 2024), accompanied by detailed documentation of what was used and how (Resnik and Hosseini 2025).

4.3. Limitations

Our evaluation was limited to English‐language journal articles indexed in PubMed and Embase, which may constrain generalizability. These databases tend to be biased toward literature from high‐income settings and do not fully capture regional or non‐English research, nor gray‐literature sources. As a result, the chatbot's performance may differ when applied to region‐specific or broader evidence databases. Furthermore, the tool's outputs depend on the behavior of the ChatGPT‐4.0 model at the time of evaluation. Because large language models evolve rapidly, future versions may generate different search strategies even with identical prompts, raising considerations for reproducibility and versioning. Accordingly, our findings should be interpreted as reflecting the performance of a specific model configuration and database context.

The chatbot is publicly hosted on the OpenAI platform. User inputs are processed according to OpenAI's data‐handling policies, which may include temporary storage for security, safety monitoring, and service reliability. We do not have access to, nor do we store or retain any user‐entered queries or data. Nonetheless, users should avoid entering sensitive or proprietary information, consistent with common practice when using third‐party digital research tools.

Generative AI has the potential to transform the development of systematic review search strategies. In our study, the chatbot showed moderate success in retrieving relevant studies, offering support, saving time, and lowering barriers in specialized expertise and technical skills. However, it should be seen as a complement to expert‐led searching rather than a replacement. With well‐crafted prompts, the integration of controlled vocabularies, and guidance from experienced librarians, generative AI tools could significantly enhance systematic review workflows, making them faster, more accessible, and more scalable—particularly in contexts with limited resources and expertise. It is also remarkable that our approach with the chatbot is limited to searching English language papers from two databases only. While this can often be a primary kickstart to an SR, the search string generated from the chatbot should be utilized with caution and manual search for other potential sources, such as unpublished or foreign papers, should still be carried out for a thorough SR.

5. Conclusion

This study demonstrates the practical feasibility of using a custom ChatGPT‐4.0–based chatbot to support systematic review (SR) search strategy development in nursing and healthcare research, substantially reducing the time and technical burden of producing database‐ready queries. Across 50 Cochrane intervention SRs published in 2024, the chatbot‐generated PubMed and Embase strategies achieved moderate retrieval performance—identifying roughly two‐thirds to three‐quarters of eligible included studies—indicating that generative AI can provide a useful starting point, particularly for teams with limited access to expert information specialists. However, performance varied and was constrained by factors such as outcomes not appearing in abstracts, inconsistent terminology (especially in complex topics), indexing limitations, and concerns about reproducibility as model outputs may change over time. Accordingly, the chatbot should be positioned as an augmentative tool rather than a replacement for librarian‐led searching: its outputs require expert review, careful documentation, and supplementation with additional sources (e.g., non‐English literature and gray literature) to support comprehensive, transparent, and reproducible SR methods.

Funding

The authors have nothing to report.

Ethics Statement

The authors have nothing to report.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Data S1: jnu70076‐sup‐0001‐Supinfo.docx.

JNU-58-0-s001.docx (81.9KB, docx)

Acknowledgments

We would like to thank Miss Chin, Mien Chew Annelissa from Medicine+Science Library, National University of Singapore, Singapore, for providing her expert opinion to improve the chatbot.

Contributor Information

Wai San Wilson Tam, Email: nurtwsw@nus.edu.sg.

Arthur Tang, Email: arthur.tang@rmit.edu.vn.

Data Availability Statement

The data used in this study were based on published Cochrane Reviews and the full citations of the Cochrane Reviews have been listed in Data S1.

References

  1. Bramer, W. M. , de Jonge G. B., Rethlefsen M. L., Mast F., and Kleijnen J.. 2018. “A Systematic Approach to Searching: An Efficient and Complete Method to Develop Literature Searches.” Journal of the Medical Library Association 106, no. 4: 531–541. 10.5195/jmla.2018.283. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Bui, N. , Nguyen G., Nguyen N., et al. 2025. “Fine‐Tuning Large Language Models for Improved Health Communication in Low‐Resource Languages.” Computer Methods and Programs in Biomedicine 263: 108655. 10.1016/j.cmpb.2025.108655. [DOI] [PubMed] [Google Scholar]
  3. Bullers, K. , Howard A. M., Hanson A., et al. 2018. “It Takes Longer Than You Think: Librarian Time Spent on Systematic Review Tasks.” Journal of the Medical Library Association 106, no. 2: 198–207. 10.5195/jmla.2018.323. [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Cheng, K. K. F. , Lim Y. T. E., Koh Z. M., and Tam W. W. S.. 2017. “Home‐Based Multidimensional Survivorship Programmes for Breast Cancer Survivors.” Cochrane Database of Systematic Reviews 8, no. 8: CD011152. 10.1002/14651858.CD011152.pub2. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Choi, B. K. , Verbeek J. H., Tam W. W., and Jiang J. Y.. 2010. “Exercises for Prevention of Recurrences of Low‐Back Pain.” Cochrane Database of Systematic Reviews 2010, no. 1: CD006555. 10.1002/14651858.CD006555.pub2. [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Erwin, P. J. 2004. “By the Clock: How Much Time Does an Expert Search Take? [Expert Searching].” MLA News 370: 1–12 [Google Scholar]. [Google Scholar]
  7. Glasziou, P. , Irwig L., Bain C., and Colditz G.. 2001. Systematic Reviews in Health Care a Practical Guide, 148. Cambridge University Press. [Google Scholar]
  8. Gosak, L. , Štiglic G., Pruinelli L., and Vrbnjak D.. 2025. “PICOT Questions and Search Strategies Formulation: A Novel Approach Using Artificial Intelligence Automation.” Journal of Nursing Scholarship 57, no. 1: 5–16. 10.1111/jnu.13036. [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. GovTech . 2023. The CO‐STAR Approach to Writing Your Prompts, Prompt Engineering Playbook. GovTech Data Science & AI Division. https://www.developer.tech.gov.sg/products/collections/data‐science‐and‐artificial‐intelligence/playbooks/prompt‐engineering‐playbook‐beta‐v3.pdf. [Google Scholar]
  10. Inita . 2024. “How to Create an AI Assistant With ChatGPT 4.0.” Accessed March 10, 2025. https://smb.inita.com/blog/how‐to‐create‐an‐ai‐assistant‐with‐chatgpt‐4‐0/.
  11. Kwok, K. O. , Huynh T., Wei W. I., Wong S. Y. S., Riley S., and Tang A.. 2024. “Utilizing Large Language Models in Infectious Disease Transmission Modelling for Public Health Preparedness.” Computational and Structural Biotechnology Journal 23: 3254–3257. 10.1016/j.csbj.2024.08.006. [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Lax, S. J. , Van Vogt E., Candy B., et al. 2024. “Topical Anti‐Inflammatory Treatments for Eczema: Network Meta‐Analysis.” Cochrane Database of Systematic Reviews 8, no. 8: CD015064. 10.1002/14651858.CD015064.pub2. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Lefebvre, C. , Manheimer E., and Glanville J.. 2008. “Chapter 6 Searching for Studies.” In Cochrane Handbook for Systematic Reviews of Interventions: Cochrane Book Series, edited by Higgins J. P. T. and Green S.. Cochrane Collaboration. [Google Scholar]
  14. Ngai, S. P. , Jones A. Y., and Tam W. W.. 2016. “Tai Chi for Chronic Obstructive Pulmonary Disease (COPD).” Cochrane Database of Systematic Reviews 2016, no. 6: CD009953. 10.1002/14651858.CD009953.pub2. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Rao, S. , Esvaran M., Chen L., et al. 2022. “Probiotic supplementation in neonates with congenital gastrointestinal surgical conditions: a pilot randomised controlled trial.” Pediatric Research 92, no. 4: 1122–1131. 10.1038/s41390-021-01884-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Resnik, D. B. , and Hosseini M.. 2025. “The Ethics of Using Artificial Intelligence in Scientific Research: New Guidance Needed for a New Tool.” AI and Ethics 5, no. 2: 1499–1521. 10.1007/s43681-024-00493-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Saleh, A. A. , Ratajeski M. A., and Bertolet M.. 2014. “Grey Literature Searching for Health Sciences Systematic Reviews: A Prospective Study of Time Spent and Resources Utilized.” Evidence Based Library and Information Practice 9, no. 3: 28–50. 10.18438/B8DW3K. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Stiglic, G. , Gosak L., San W. T., and Vrbnjak D.. 2025. “Human‐Led and Artificial Intelligence‐Automated Critical Appraisal of Systematic Reviews: Comparative Evaluation.” Nurse Education in Practice 89: 104614. 10.1016/j.nepr.2025.104614. [DOI] [PubMed] [Google Scholar]
  19. Sutton, A. J. , Cooper N. J., and Jones D. R.. 2009. “Evidence Synthesis as the Key to More Coherent and Efficient Research.” BMC Medical Research Methodology 9: 29. 10.1186/1471-2288-9-29. [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Tam, W. , Huynh T., Tang A., Luong S., Khatri Y., and Zhou W.. 2023. “Nursing Education in the Age of Artificial Intelligence Powered Chatbots (AI‐Chatbots): Are We Ready Yet?” Nurse Education Today 129: 105917. 10.1016/j.nedt.2023.105917. [DOI] [PubMed] [Google Scholar]
  21. Tang, A. , Ho R., Yu R., et al. 2023. “Can Artificial Intelligence Help Us Overcome Challenges in Geriatrics?” Geriatric Nursing (New York, N.Y.) 52: A1–A2. 10.1016/j.gerinurse.2023.06.007. [DOI] [PubMed] [Google Scholar]
  22. Tang, A. , Li K. K., Kwok K. O., Cao L., Luong S., and Tam W.. 2024. “The Importance of Transparency: Declaring the Use of Generative Artificial Intelligence (AI) in Academic Writing.” Journal of Nursing Scholarship 56, no. 2: 314–318. 10.1111/jnu.12938. [DOI] [PubMed] [Google Scholar]
  23. Tran, L. D. , Tung N., Macalinga E. T., Tang A., Woo B., and Tam W.. 2024. “Visual Narratives in Nursing Education: A Generative Artificial Intelligence Approach.” Nurse Education in Practice 79: 104079. 10.1016/j.nepr.2024.104079. [DOI] [PubMed] [Google Scholar]
  24. Trivedi, A. , Teo E., and Walker K. S.. 2024. “Probiotics for the Postoperative Management of Term Neonates After Gastrointestinal Surgery.” Cochrane Database of Systematic Reviews 1, no. 1: CD012265. 10.1002/14651858.CD012265.pub2. [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Wei, W. I. , Leung C. L. K., Tang A., McNeil E. B., Wong S. Y. S., and Kwok K. O.. 2024. “Extracting Symptoms From Free‐Text Responses Using ChatGPT Among COVID‐19 Cases in Hong Kong.” Clinical Microbiology and Infection 30, no. 1: 142.e1–e3. 10.1016/j.cmi.2023.11.002. [DOI] [PubMed] [Google Scholar]
  26. Woo, B. , Huynh T., Tang A., Bui N., Nguyen G., and Tam W.. 2024. “Transforming Nursing With Large Language Models: From Concept to Practice.” European Journal of Cardiovascular Nursing 23, no. 5: 549–552. 10.1093/eurjcn/zvad120. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Data S1: jnu70076‐sup‐0001‐Supinfo.docx.

JNU-58-0-s001.docx (81.9KB, docx)

Data Availability Statement

The data used in this study were based on published Cochrane Reviews and the full citations of the Cochrane Reviews have been listed in Data S1.

Articles from Journal of Nursing Scholarship are provided here courtesy of Wiley

RESOURCES