Clinical outcomes, patient satisfaction and operational efficiency of AI-powered chatbots in medicine and healthcare: protocol for an AI-aided scoping review.

Clinical outcomes, patient satisfaction and operational efficiency of AI-powered chatbots in medicine and healthcare: protocol for an AI-aided scoping review - PMC Skip to main content An official website of the United States government Here's how you know Here's how you know Search Log in Primary site navigation Search Log in Search PMC Full-Text Archive Search in PMC Journal List User Guide As a library, NLM provides access to scientific literature. Inclusion in an NLM database does not imply endorsement of, or agreement with, the contents by NLM or the National Institutes of Health. Learn more: PMC Disclaimer | PMC Copyright Notice BMJ Open . 2026 Jun 3;16(6):e113547. doi: 10.1136/bmjopen-2025-113547 Clinical outcomes, patient satisfaction and operational efficiency of AI-powered chatbots in medicine and healthcare: protocol for an AI-aided scoping review Ellen Goossens 1, 2, 3, ✉ , Pamela Mfouth Kemajou 4 , Pascal Coorevits 4 , Sebastian Probst 5, 6, 7, 8 , Dimitri Beeckman 1, 9 Author information Article notes Copyright and License information 1 University Centre for Nursing and Midwifery, Department of Public Health and Primary Care, Faculty of Medicine and Health Sciences, Ghent University, Ghent, Belgium 2 Research Centre 360° Care and Well-being, HOGENT University of Applied Sciences and Arts, Ghent, Belgium 3 School of Healthcare, HOGENT University of Applied Sciences and Arts, Ghent, Belgium 4 Unit of Medical Informatics and Statistics, Department of Public Health and Primary Care, Faculty of Medicine and Health Sciences, Ghent University, Ghent, Belgium 5 Geneva School of Health Science, HES-SO University of Applied Sciences and Arts, Geneva, Western Switserland, Switzerland 6 Care Directorate, Geneva University Hospitals, Geneva, Switzerland 7 Faculty of Medicine, University of Geneva, Geneva, Switzerland 8 Faculty of Medicine Nursing and Health Sciences, Monash University, Melbourne, Victoria, Australia 9 Swedish Centre for Skin and Wound Research (SCENTR), School of Health Sciences, Faculty of Medicine and Health, Örebro University, Örebro, Sweden Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise. None declared. ✉ Ellen Goossens; ellegoos.goossens@ugent.be Received 2025 Nov 7; Accepted 2026 May 14; Collection date 2026. Copyright © Author(s) (or their employer(s)) 2026. Re-use permitted under CC BY-NC. No commercial re-use. See rights and permissions. Published by BMJ Group. This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: https://creativecommons.org/licenses/by-nc/4.0/ . PMC Copyright notice PMCID: PMC13239711  PMID: 42236098 Abstract Abstract Introduction Artificial intelligence (AI)-powered chatbots are increasingly integrated into healthcare to support administrative processes, health education and chronic disease management. These systems simulate human dialogue through natural language processing and machine learning, enabling dynamic and context-aware interactions. Despite their rapid adoption, there is limited synthesis of existing research describing how these technologies are applied across different healthcare contexts and what outcomes have been reported. This scoping review aims to map and describe the existing literature on the use of AI-powered chatbots in healthcare with a focus on clinical outcomes, patient satisfaction and operational efficiency. It will identify the types of studies conducted, their key characteristics and existing research gaps to guide future research. Methods and analysis Following the Joanna Briggs Institute methodology and Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews guidelines, a comprehensive search will be conducted across Medline (PubMed), CINAHL, Embase, Web of Science, The Cochrane Library and MedRxiv from database inception to 10 September 2025. Studies published in English, French, Dutch or German, involving AI-powered chatbots in any healthcare context reporting on clinical outcomes and/or patient satisfaction and/or operational efficiency will be included. Studies without full-text availability, protocols, trial registrations, reviews and studies conducted solely in educational settings will be excluded. Title and abstract screening will be supported by ASReview LAB, an AI-based active learning tool to enhance efficiency. Screening and data extraction will be conducted independently by two reviewers with disagreements resolved by a third reviewer. Findings will be synthesised narratively and presented using structured evidence tables categorised by chatbot type, clinical healthcare context and reported outcomes. Ethics and dissemination Ethical approval is not required, as this study involves the analysis of published data only. The results of this scoping review will be disseminated through publication in a peer-reviewed journal, presentations at academic conferences and established professional networks. Trial registration number Open Science Framework (OSF), https://doi.org/10.17605/OSF.IO/8UE3B . Keywords: Artificial Intelligence, Patient Satisfaction, Medicine, Health Care Costs, Patient Reported Outcome Measures, Treatment Outcome STRENGTHS AND LIMITATIONS OF THIS STUDY. This scoping review aims to provide a comprehensive mapping of artificial intelligence (AI)-powered chatbot applications across healthcare settings, focusing on clinical outcomes, patient satisfaction and operational efficiency. The review follows established methodological guidance from the Joanna Briggs Institute and Preferred Reporting Items for Systematic reviews and Meta-Analyses extension for Scoping Reviews, ensuring transparency and reproducibility. The integration of AI-assisted screening enhances efficiency but introduces a risk of missing relevant studies (false negatives), which is mitigated through safeguards including predefined stopping criteria, the use of landmark studies, iterative model training and validation and independent dual human review. Fragmentation in study designs, populations and outcome measures may limit comparability and interpretation of findings. Given the rapidly expanding body of literature in this field, some ongoing or yet-to-be-published studies may not be captured in this review. Introduction Rationale The digital transformation of healthcare accelerated the integration of artificial intelligence (AI) into clinical and organisational processes. 1 Among these innovative digital tools, AI-powered chatbots - also referred to as conversational agents - have emerged as versatile tools capable of simulating human conversation through text, speech or both, 2 , 7 using natural language processing and machine learning. 1 , 36 8 Their applications in healthcare range from administrative support 9 - such as appointment scheduling, sending reminders, 1 10 disseminating healthcare information 1 9 11 and managing electronic health records 9 11 - to more complex roles including health education, 7 9 care management, 9 12 behaviour change promotion, 9 emotional support 1 7 and chronic disease management. 12 Recent advances in generative AI, particularly following the release of systems such as OpenAI’s ChatGPT in 2022, have expanded the capabilities, accessibility and real-world applicability of chatbot technologies. 13 This shift from rule-based and retrieval-based systems towards generative models has contributed to a rapidly growing and increasingly fragmented body of literature across diverse healthcare settings and use cases. 4 14 15 Despite this growth, the current evidence remains fragmented and dispersed across disciplines, clinical settings and study designs. Studies vary widely in designs, populations, settings and reported outcomes, while also highlighting challenges related to accuracy, reliability, bias and ethical considerations such as privacy and misinformation. 19 16 , 18 This variability makes it difficult to obtain a comprehensive understanding of how AI-powered chatbots influence healthcare delivery. Existing reviews have provided valuable insights, but are often limited in scope. Many focus on specific clinical domains (eg, mental health, 19 20 oncology 21 , 24 or reproductive health 25 26 ) or on isolated aspects such as usability, acceptability or technical performance, 7 9 27 rather than offering a mapped overview across healthcare contexts. In addition, several reviews predate the widespread adoption of generative AI technologies, limiting their relevance to the current landscape. 13 Importantly, the impact of AI-powered chatbots in healthcare is inherently multidimensional. It spans clinical outcomes (eg, health status and treatment effects), 16 28 patient satisfaction (eg, user experience and perceived quality of care) 29 30 and operational efficiency (eg, workflow optimisation and resource utilisation). 31 These domains represent complementary dimensions of healthcare evaluation and are widely used in health services research to assess interventions across clinical effectiveness, patient experience and system-level performance. 32 33 To date, no scoping review has yet mapped the existing literature on AI-powered chatbots in healthcare across these three domains while accounting for recent advances in generative AI. A comprehensive overview of the literature is therefore needed to clarify how these technologies are currently applied, to identify patterns and gaps in the evidence, and to inform future research and implementation. A scoping review is particularly appropriate for this purpose, as it allows for a comprehensive and structured mapping of emerging, fragmented evidence across a broad range of study designs and outcomes. Research question What is known from existing literature about the use of AI-powered chatbots in medicine and healthcare, particularly in relation to clinical outcomes, patient satisfaction and operational efficiency? Objectives The main objective of the scoping review is to map and describe the existing evidence on the application of AI-powered chatbots in healthcare and describe their reported associations with clinical outcomes, patient satisfaction and operational efficiency. Specifically, the objectives of this scoping review protocol are: To identify studies that explore the use, application or evaluation of AI-powered chatbots in any healthcare context. To describe the types, characteristics and methodological features of the studies identified, including target populations, healthcare settings and chatbot functionalities. To extract and synthesise data narratively to identify overarching themes related to clinical practice, patient experience and operational processes. Methods and analysis This scoping review protocol follows the Preferred Reporting Items for Systematic Review and Meta-Analysis Protocols (PRISMA-P) guidelines. 28 The scoping review will be conducted in accordance with the Joanna Briggs Institute (JBI) methodology for scoping reviews. 29 The review is scheduled to be conducted on 10 September 2025. Eligibility criteria Studies will be selected through a systematic process involving database searches and screening against predefined inclusion and exclusion criteria. The selection will be performed by two reviewers, and discrepancies will be resolved through consensus or consultation with a third reviewer. The inclusion criteria are structured following the Population–Concept–Context framework recommended by JBI to clearly define eligible participants, interventions and settings. 29 The following inclusion criteria apply: Population (P): No restrictions on participant characteristics (eg, age, gender or health status). Concept (C): Studies involving AI-powered chatbots, including rule-based, retrieval-based and generative-based models. Studies reporting on clinical outcomes, and/or patient satisfaction and/or operational efficiency. Context (C): Studies conducted within the domains of healthcare and/or medicine. There will be no restrictions based on publication date or geographical regions. Due to the language skills of the reviewers, the review will only include articles published in English, French, Dutch or German. Records will be excluded if they: Are solely available in abstract form without access to the full text. Constitute study protocols or trial registrations (since no results are available yet) or reviews. They are conducted in educational settings. Information sources The databases Medline (PubMed), CINAHL (EBSCOhost), Embase, Web of Science and The Cochrane Library, as well as preprints from recognised repositories (eg, medRxiv), will be searched to identify relevant records. The database search, conducted on 10 September 2025, covered all publications from inception to that date. All retrieved records will be imported into Rayyan for deduplication. Search strategy The search strategy is developed in collaboration with an information specialist from the Knowledge Centre for Health in Ghent. It was initially designed for the Medline (PubMed) search engine and subsequently adapted for use in the other aforementioned databases. Preprints will also be screened and included if they meet the eligibility criteria. The strategy combines key terms (eg, MeSH terms such as Artificial Intelligence, patient satisfaction) with free-text keywords, using Boolean operators. The complete search string and the adapted versions for each database are presented in table 1 and in the online supplemental material , respectively. The search strategy is structured around three main concepts: Artificial Intelligence (eg, AI-driven, AI-generated), chatbots (eg, conversational agent) and outcomes of interest (eg, clinical outcome, patient satisfaction and efficiency). Table 1. Search strategy for the Medline (PubMed) database. Name of database (interface): Medline (PubMed) Concept Line number Search strategy Concept 1: Artificial Intelligence 1 “Artificial Intelligence”[Mesh] OR “Artificial Intelligence”[Title/Abstract] OR “Generative Artificial Intelligence*”[Title/Abstract] OR GenAI*[Title/Abstract] OR “Gen-AI*”[Title/Abstract] OR “Generative AI”[Title/Abstract] OR AI[Title/Abstract] OR “AI-driven*”[Title/Abstract] OR “AI-powered*”[Title/Abstract] OR “AI-based*”[Title/Abstract] OR “AI-generated*”[Title/Abstract] OR “AI-assisted*”[Title/Abstract] OR “generative-based*”[Title/Abstract] OR “generativebased*“[Title/Abstract] OR “retrieval-based*”[Title/Abstract] OR “retrievalbased*“[Title/Abstract] OR “rule-based*”[Title/Abstract] OR “rulebased*”[Title/Abstract] Concept 2: Chatbot 2 chatbot*[Title/Abstract] OR “chat bot*”[Title/Abstract] OR “chat interface*“[Title/Abstract] OR chatmate*[Title/Abstract] OR “conversational agent*”[Title/Abstract] OR “conversational AI*”[Title/Abstract] OR “conversational bot*”[Title/Abstract] OR “conversational interface*”[Title/Abstract] OR “conversational assistant*”[Title/Abstract] OR “dialogue agent*”[Title/Abstract] OR “dialogue system*”[Title/Abstract] OR “social bot*”[Title/Abstract] OR “interactive agent*”[Title/Abstract] OR “therapy bot*”[Title/Abstract] OR “softbot*”[Title/Abstract] OR “virtual agent*”[Title/Abstract] OR “automated agent*”[Title/Abstract] OR “relational agent*”[Title/Abstract] OR “virtual coach*”[Title/Abstract] OR “virtual assistant*”[Title/Abstract] OR “digital assistant*”[Title/Abstract] OR ChatGPT*[Title/Abstract] OR Chat-GPT*[Title/Abstract] OR GPT[Title/Abstract] OR Med-PaLM[Title/Abstract] OR BioGPT[Title/Abstract] OR GatorTron*[Title/Abstract] OR “Woebot Health”[Title/Abstract] OR DeepSeek[Title/Abstract] OR Gemini[Title/Abstract] OR Perplexity[Title/Abstract] OR Claude[Title/Abstract] OR “Microsoft Copilot”[Title/Abstract] OR Grok[Title/Abstract] OR “Meta AI”[Title/Abstract] OR Mistral[Title/Abstract] OR Abridge[Title/Abstract] OR Replika[Title/Abstract] OR Wysa[Title/Abstract] OR Tess[Title/Abstract] OR Youper[Title/Abstract] OR “Ada Health”[Title/Abstract] Concept 3: Outcomes of interest 3 “health care quality, access, and evaluation”[MeSH]) OR “quality of health care”[Mesh] OR “functional status”[Mesh] OR “health status”[Mesh] OR “health status indicators”[MeSH] OR “treatment adherence and compliance”[Mesh] OR “patient satisfaction”[MeSH] OR “quality of life”[Mesh] OR efficiency[MeSH] OR workflow[Mesh] OR “costs and cost analysis”[MeSH] OR “outcome assessment*”[Title/Abstract] OR “process assessment*”[Title/Abstract] OR “patient-outcome assessment*”[Title/Abstract] OR “ clinical outcome *“[Title/Abstract] OR “patient-reported outcome*”[Title/Abstract] OR “therapeutic outcome*“[Title/Abstract] OR “health outcome*“[Title/Abstract] OR “health status*“[Title/Abstract] OR “health indicator*”[Title/Abstract] OR “level of health*”[Title/Abstract] OR “health level*”[Title/Abstract] OR “general health*“[Title/Abstract] OR “overall health*“[Title/Abstract] OR “health risk*“[Title/Abstract] OR “functional status*“[Title/Abstract] OR “treatment outcome*“[Title/Abstract] OR “clinical benefit*“[Title/Abstract] OR “clinical useful*”[Title/Abstract] OR “clinical impact*“[Title/Abstract] OR “symptom improvement*“[Title/Abstract] OR “symptom control*“[Title/Abstract] OR “symptom management”[Title/Abstract] OR “disease management*“[Title/Abstract] OR “disease control*“[Title/Abstract] OR “treatment adherence*“[Title/Abstract] OR “treatment compliance*“[Title/Abstract] OR “medication adherence*“[Title/Abstract] OR “drug adherence*“[Title/Abstract] OR “drug non-adheren*“[Title/Abstract] OR “drug nonadheren*“[Title/Abstract] OR “therapeutic adherence*“[Title/Abstract] OR “patient adherence*“[Title/Abstract] OR “patient non-adheren*“[Title/Abstract] OR “patient nonadheren*“[Title/Abstract] OR “medication nonadherence*“[Title/Abstract] OR “medication non-adherence*“[Title/Abstract] OR “medication compliance*“[Title/Abstract] OR “patient compliance*“[Title/Abstract] OR “medication noncompliance*“[Title/Abstract] OR “medication non-compliance*“[Title/Abstract] OR “drug compliance*“[Title/Abstract] OR “quality of health care*“[Title/Abstract] OR “quality of healthcare”[Title/Abstract] OR “quality of care*“[Title/Abstract] OR “care quality”[Title/Abstract] OR “healthcare quality”[Title/Abstract] OR “health care quality”[Title/Abstract] OR “healthcare access*“[Title/Abstract] OR “health care access*"[Title/Abstract] OR “ patient satisfaction* “[Title/Abstract] OR “patient’s satisfaction*”[Title/Abstract] OR “patients’ satisfaction*”[Title/Abstract] OR “user satisfaction”[Title/Abstract] OR “user-friendliness”[Title/Abstract] OR “user friendly”[Title/Abstract] OR “patient support*”[Title/Abstract] OR “satisfactory response*”[Title/Abstract] OR “satisfaction level*”[Title/Abstract] OR “quality of life*“[Title/Abstract] OR QoL[Title/Abstract] OR HRQOL*[Title/Abstract] OR “life quality”[Title/Abstract] OR “PSQ-18”[Title/Abstract] OR “Patient Satisfaction Questionnaire Short Form”[Title/Abstract] OR “Patient Satisfaction Questionnaire-18”[Title/Abstract] OR “CAHPS*”[Title/Abstract] OR “Consumer Assessment of Healthcare Providers and Systems”[Title/Abstract] OR “PPE-15”[Title/Abstract] OR “Picker Patient Experience Questionnaire”[Title/Abstract] OR “patient experience*”[Title/Abstract] OR “patient perception*”[Title/Abstract] OR “patient-reported satisfaction”[Title/Abstract] OR “service satisfaction”[Title/Abstract] OR “satisfaction rate*”[Title/Abstract] OR “perceived quality”[Title/Abstract] OR “PREMs*“[Title/Abstract] OR “patient-reported experience*”[Title/Abstract] OR “user experience*“[Title/Abstract] OR “CSQ-8*“[Title/Abstract] OR “Client Satisfaction Questionnaire-8”[Title/Abstract] OR “user experience questionnaire*“[Title/Abstract] OR “patient reported outcome measure*“[Title/Abstract] OR PROMs*[Title/Abstract] OR “patient centered outcome*“[Title/Abstract] OR “patient preference*“[Title/Abstract] OR “patient need*“[Title/Abstract] OR “patient expectation*“[Title/Abstract] OR efficient* [Title/Abstract] OR inefficient*[Title/Abstract] OR efficienc*[Title/Abstract] OR inefficienc*[Title/Abstract] OR effectiveness*[Title/Abstract] OR efficacy*[Title/Abstract] OR effective*[Title/Abstract] OR “time efficiency”[Title/Abstract] OR “costs and cost analysis”[Title/Abstract] OR “cost analys*“[Title/Abstract] OR “cost measure*“[Title/Abstract] OR “cost*“[Title/Abstract] OR “health economic*“[Title/Abstract] OR “cost-effective*“[Title/Abstract] OR “cost efficien*“[Title/Abstract] OR “cost benefit analys*“[Title/Abstract] OR “consultation time*“[Title/Abstract] OR “resource utilization”[Title/Abstract] OR “workflow*“[Title/Abstract] OR “work flow*“[Title/Abstract] OR “workload*“[Title/Abstract] OR “healthcare productivity”[Title/Abstract] OR “organizational productivity”[Title/Abstract] OR “organisational productivity”[Title/Abstract] OR accessibility*[Title/Abstract] OR usability[Title/Abstract] OR “system usability scale”[Title/Abstract] OR “SUS”[Title/Abstract] OR “useful*“[Title/Abstract] OR “ease of use”[Title/Abstract] OR “acceptab*“[Title/Abstract] OR “acceptance*“[Title/Abstract] OR “Net Promoter Score*“[Title/Abstract] OR safe*[Title/Abstract] OR “diagnostic precision”[Title/Abstract] OR accuracy*[Title/Abstract] Combination of concepts 4 1, 2, 3 AND Open in a new tab AI, artificial intelligence. Data management Selection process Following the database search, all retrieved references will be imported into the reference management software EndNote 2025 (V.22) for initial data storage. 30 All retrieved records will then be transferred to Rayyan, an AI-powered systematic review management platform, for an initial round of deduplication. 31 Deduplication is particularly critical in AI-aided screening, as duplicate records can disproportionately influence the AI model’s classification algorithm. 32 Rayyan’s built-in duplicate detection tool, supplemented by manual review, will be used to identify and remove duplicates. 31 33 Among available tools, Rayyan has demonstrated the highest sensitivity for deduplication and offers an average time saving of 40%. Title and abstract screening will be supported by ASReview LAB (V.2.0), an AI tool that applies active learning to prioritise the most relevant records for human review. 34 35 Active learning allows the model to iteratively select which documents to present for labelling, thereby reducing the manual screening burden while maintaining methodological rigour. 36 In ASReview LAB, these labels are binary (eg, relevant or irrelevant) and are assigned to an article’s title and abstract. 35 A PRISMA flow diagram will be used to report the study selection process in the final review. Step 1: preparatory screening and model training Two independent reviewers will begin by screening a random sample of 1% (177 of 17 700 articles) to establish a shared understanding of the inclusion and exclusion criteria. Following this preparatory screening, discrepancies will be discussed to reach a consensus between the two reviewers. If consensus cannot be reached, a third reviewer will be consulted. This step also enables a preliminary estimate of the total number of relevant records by calculating the proportion of relevant records in the sample and extrapolating to the full dataset. 37 This estimate will inform the stopping criterion for the AI-assisted screening. Step 2: AI-aided screening The full dataset will be uploaded into two locally run instances of ASReview LAB, where it will be independently screened by two reviewers. The consensus labels from the preparatory step will serve as training data for the active learning model. Minimum one article has to be labelled as relevant and one labelled as irrelevant in ASReview LAB. 35 The model will be configured using ELAS u4 with the default settings: (1) Feature extractor: Term Frequency-Inverse Document Frequency, (2) Classifier: Support Vector Machine, (3) Sampling strategy: balanced sample weight resampling and (4) Query strategy: maximum certainty. These settings have demonstrated strong performance across multiple datasets. 38 The model will iteratively present records for labelling, retraining after each decision to improve its predictions. 38 A stopping criterion is used to determine when to end the active learning process in screening software. Screening will continue until the following SAFE stopping criteria are met: (1) All predefined key (landmark) articles have been presented by the model, (2) At least twice the estimated number of relevant articles has been screened, (3) A minimum of 10% of the dataset has been reviewed and (4) No relevant articles have been identified in the last 100 screened records. 37 Before starting the screening process, the research team will first put together a list of key articles to guide the review. Once both reviewers independently meet the stopping criterion, each reviewer will review the records suggested only to the other by the AI model. Any remaining discrepancies will be resolved through discussion. If consensus cannot be reached, a third reviewer will be consulted. At this phase, Reviewer 1 will undertake a second round of AI-aided screening of titles and abstracts using a more complex and customised model. This model will employ mpnet Sentence Bidirectional Encoder Representations from Transformers as the feature extractor and a random forest classifier with balanced sample weighting and a mixed querying strategy (95% maximum-certainty and 5% uncertainty). 38 It is generally advisable to commence screening with a simpler model and advance to a more computationally demanding one once a sufficient quantity of labelled data has been accumulated, as complex models typically require larger training sets to achieve optimal performance. 39 Accordingly, the consensus decisions on inclusion and exclusion reached by the research team in the preceding steps 1 and 2 will serve as the training dataset for this model. The output generated from this screening process will subsequently be reviewed by Reviewer 1, ensuring that each article identified by the AI model undergoes a second human screening. The output of this screening step will then again be passed to reviewer 2 for independent screening of a randomly selected sample of the studies suggested by the model. Any discrepancies in classification between reviewers 1 and 2 will be resolved through discussion and consensus within the research team. Step 3: full-text screening All articles retained after title and abstract screening will undergo full-text review by two independent reviewers. Disagreements will be resolved through discussion or, if necessary, adjudication by a third reviewer. This ensures that every article suggested by the AI model is reviewed by at least two human reviewers. To ensure the review remains up to date, an email alert will be set up to notify the first reviewer of newly published studies during the review period. These articles will be screened using the same process. Data collection process Data will be extracted using Elicit, a free online AI research assistant designed to automate elements of researchers’ workflows related to evidence synthesis and text extraction. 40 , 42 Elicit will facilitate the data collection process by generating a preliminary data extraction table. A standardised data extraction form will be used to capture the key characteristics and findings of each included study. The first reviewer will extract data using Elicit and verify the preliminary extraction table. Subsequently, the second reviewer will validate the extracted information. All extracted data will be validated by at least one human reviewer; in this study, validation will be performed by two reviewers. To ensure the extraction form is comprehensive and aligned with the review objectives, it will be piloted using a sample of six studies. Any necessary adjustments will be made based on this pilot. Any discrepancies will be resolved through discussion, with a third reviewer consulted if needed. In cases where data are missing or unclear, the corresponding authors will be contacted via email. An initial request will be followed by a second reminder after 4 weeks if no response is received. If no reply is obtained after two attempts, the data will be considered unobtainable. The search and selection process will be summarised in a PRISMA flowchart. Data items An evidence table will be developed to ensure comprehensive and reliable data extraction. The following items will be extracted: Study characteristics: Author(s). Year of publication. Country. Study design. Study duration. Study setting and sample: Sample size. Study setting (eg, hospital, primary care). Domain (eg, medicine, nursing, physiotherapy, mental health and social care). Specialisation (eg, geriatrics, cardiology, radiology, oncology and psychiatry). Participant roles (eg, patients, caregivers, healthcare providers and administrative staff). Age of participants. Gender of participants. Condition of participants. Professional groups (eg, general practitioners, nurses, physiotherapists and psychologists). Geographical location. Eligibility criteria. Intervention details: Function of the AI-powered chatbot (eg, diagnosis support, health education, mental health counselling and administrative tasks). Criteria for AI-powered chatbot implementation. Description of the intervention group and intervention (if applicable). Description of the control group (if applicable). Duration and frequency of AI-powered chatbot use in the study. User engagement and adherence rates (if reported). Outcomes, along with their definitions and time points, will be categorised into (but not limited to) the following areas: Clinical outcomes (eg, symptom improvement, disease management and medication adherence). Patient satisfaction (eg, user experience). Operational efficiency (eg, reduction in consultation time, healthcare resource utilisation and cost-effectiveness). Results: Summary data for each reported outcome. An estimate of the effect size, including p value (if available). Methodological considerations. Key conclusions – summary of the study authors’ main findings, implications and limitations. Outcomes and prioritisation The three outcome domains - clinical outcomes, patient satisfaction and operational efficiency - were selected to reflect key and complementary dimensions of healthcare evaluation. These domains represent interconnected levels of health system impact, ranging from patient-level experiences to clinical effectiveness and system-level performance. Their selection is grounded in health services research, where digital health interventions are commonly evaluated across clinical effectiveness, patient experience and organisational performance. 43 44 The following outcomes will be considered in this scoping review: Clinical outcomes: clinical outcome is defined as measurable changes in patient’s symptoms, overall health status, functional ability and quality of life resulting from healthcare interventions. 45 In the context of AI-powered chatbots, these may include outcomes such as symptom improvement, disease management or treatment adherence. 16 46 Patient satisfaction: patient satisfaction is defined as the degree to which patients perceive the provided information, communication and overall care experience as meeting their expectations, needs and preferences. 47 This includes aspects such as usability, trust, perceived usefulness and user experience. 48 49 Operational efficiency: operational efficiency in healthcare is defined as the ability of the healthcare system to deliver services effectively, measured by the alignment between planned and actual healthcare services provided. It reflects how well resources are used to ensure timely, cost-effective and high-quality care. 50 In studies of AI-powered chatbots, this may include outcomes such as reduced workload, improved workflow or resource utilisation. 51 In addition, studies will be included if they report on at least one of the three predefined domains, to ensure comprehensive coverage of the evidence base. Any additional outcomes reported in the included studies will also be documented to support a broad mapping of the literature. By examining clinical, patient-reported and operational outcomes, this review aims to provide a more integrated understanding of the role and impact of AI-powered chatbots in healthcare systems. Data synthesis Data extraction will be conducted as outlined above. A narrative and descriptive synthesis will be employed to provide a comprehensive overview of the findings from the included studies. Extracted data will be categorised according to key themes and outcome domains. Each theme will be summarised descriptively, highlighting methodologies, results and conclusions. The synthesis will explore both commonalities and differences across studies to identify patterns, gaps and inconsistencies in the evidence base. Variability in study design, intervention characteristics and outcome measures will be examined to assess fragmentation and its potential influence on findings. Finally, the synthesis can serve as a framework for clinical practice and inform clinical decision-making. 52 Patient and public involvement None. Ethics and dissemination Ethical approval is not required for this scoping review, as it involves the analysis of previously published studies. No primary data collection or access to patient data will be conducted. The findings of this scoping review will be reported following the PRISMA extension for Scoping Reviews guidelines to ensure transparency and methodological rigour. The protocol has been registered in the Open Science Framework (OSF) and is available at https://doi.org/10.17605/OSF.IO/8UE3B to ensure transparency and replicability. The results of this scoping review will be disseminated through publication in a peer-reviewed journal, presentations at academic conferences and established professional networks. Authorship will be determined based on substantial contributions to the study and manuscript development. To qualify for authorship, an individual must meet the following criteria: Significant contributions to the study design, data collection, analysis or interpretation of results. Active involvement in drafting or critically revising the manuscript. Final approval of the version to be submitted and agreement to be accountable for all aspects of the work. The first author will prepare the initial draft of the manuscript summarising the review findings. All co-authors will review and approve the final version prior to submission to a peer-reviewed journal. Discussion AI-powered chatbots are widely integrated into healthcare systems to streamline workflows and enhance service delivery. 49 53 , 55 Their applications range from administrative support to direct patient care, with growing research describing their use across various clinical contexts. 1 10 However, despite their expanding role, no scoping review to date has yet summarised the current evidence on the use of AI-powered chatbots in relation to clinical outcomes, patient satisfaction and operational efficiency. Understanding this evidence base is essential to map the current evidence landscape, to clarify how chatbots are being applied in practice and to highlight areas requiring further research. This review aims to provide an overview of existing research, identify thematic trends and highlight gaps in knowledge regarding AI-powered chatbot applications in healthcare and medicine. Existing studies are dispersed across disciplines and settings, and use cases, making it difficult to obtain a cohesive overview of current trends, applications and knowledge gaps. The planned scoping review is therefore timely and necessary to synthesise the breadth and scope of research in this area and to inform future developments in digital healthcare. By identifying gaps in existing studies, this review will inform future research priorities and guide the development and evaluation of AI-based interventions. Conceptual considerations and assumptions This review is based on several underlying assumptions. First, it assumes that included studies provide sufficiently detailed descriptions of AI-powered chatbot interventions to allow meaningful classification and comparison. Second, it assumes that reported outcomes - such as clinical outcomes, patient satisfaction and operational efficiency - can be meaningfully interpreted across fragmented study designs and contexts. Ethical and practical considerations Beyond methodological considerations, the implementation of AI-powered chatbots in healthcare raises important ethical and practical challenges. Technical limitations - such as inaccuracies, hallucinations and limited contextual understanding – may affect the reliability of chatbot-generated information. In parallel, ethical concerns include bias, data privacy, informed consent and the potential of misinformation. 16 , 18 These challenges are particularly relevant in clinical settings, where inappropriate or biased outputs may negatively affect patient safety, autonomy and trust in healthcare systems. Recent research also highlights the importance of human-computer collaboration, emphasising that AI systems should support rather than replace clinical decision-making. 17 18 Enhancing transparency, explainability and appropriate human oversight is therefore essential to ensure safe and responsible implementation of AI-powered chatbots in healthcare. Methodological considerations In addition to mapping the evidence, this protocol contributes to methodological innovation by incorporating AI-assisted screening using ASReview LAB. This tool, which applies active learning to prioritise relevant records, has been shown to reduce screening workloads by an average of 60.02% in systematic reviews. 56 By decreasing the number of records requiring manual review, AI tools also help reduce labelling errors, reported to occur in up to 21% of cases among experienced reviewers and up to 58% among less experienced ones. 57 Nevertheless, the use of AI-assisted screening introduces important limitations. A key concern is the risk of false negatives, whereby relevant studies may not be identified during the initial screening process and therefore remain unavailable for subsequent human review. This limitation arises from the model’s reliance on iterative learning, which prioritises a subset of records and may result in the omission of up to 5% of relevant studies. 58 To mitigate this risk, several safeguards are incorporated into the screening process. These include predefined stopping criteria, the use of key (landmark) studies and multiple rounds of model training and validation. In addition, all records identified by the AI model will be reviewed by at least two independent human reviewers, with discrepancies resolved through discussion or adjudication by a third reviewer. A random subset of included studies will also be independently checked to further enhance reliability. Despite these measures, the possibility of missed studies cannot be fully excluded and will be considered when interpreting the findings. More broadly, these challenges reflect the ongoing balance between efficiency gains and potential risks when integrating AI into evidence synthesis workflows. 59 Furthermore, AI-assisted approaches do not eliminate the potential for bias, as final inclusion decisions remain dependent on human judgment. Additionally, given the rapidly expanding body of literature in this field, some ongoing or yet-to-be-published studies may not be captured. Strengths and limitations This protocol has several strengths. It is based on established methodological guidance for scoping reviews and applies a transparent and reproducible approach to study identification, selection and data extraction. In addition, the integration of AI-assisted screening supports the feasibility of managing large and rapidly expanding bodies of literature. However, several limitations should be acknowledged. First, the use of AI-assisted screening introduces a residual risk of missing relevant studies (false negatives), despite the implementation of safeguards such as stopping criteria, landmark studies and dual human review. Second, the expected fragmentation in study designs, populations and outcome measures may limit the comparability of findings. Third, the inclusion of only selected languages may lead to the exclusion of relevant studies published in other languages. Fourth, given the rapid pace of developments in AI, newly emerging evidence may not be captured at the time of the search. Finally, as with all evidence syntheses, the findings of this review will depend on the quality, completeness and consistency of reporting in the included studies. Supplementary material online supplemental file 1 bmjopen-16-6-s001.pdf (226KB, pdf) DOI: 10.1136/bmjopen-2025-113547 Footnotes Funding: The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors. Prepublication history and additional supplemental material for this paper are available online. To view these files, please visit the journal online ( https://doi.org/10.1136/bmjopen-2025-113547 ). Provenance and peer review: Not commissioned; externally peer reviewed. Patient consent for publication: Not applicable. Patient and public involvement: Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research. References 1. Sun G, Zhou YH. AI in healthcare: navigating opportunities and challenges in digital communication. Front Digit Health . 2023;5:1291132. doi: 10.3389/fdgth.2023.1291132. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 2. Allouch M, Azaria A, Azoulay R. Conversational Agents: Goals, Technologies, Vision and Challenges. Sensors (Basel) 2021;21:8448. doi: 10.3390/s21248448. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 3. Kocaballi AB, Quiroz JC, Rezazadegan D, et al. Responses of Conversational Agents to Health and Lifestyle Prompts: Investigation of Appropriateness and Presentation Structures. J Med Internet Res. 2020;22:e15823. doi: 10.2196/15823. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 4. Kurniawan MH, Handiyani H, Nuraini T, et al. A systematic review of artificial intelligence-powered (AI-powered) chatbot intervention for managing chronic illness. Ann Med. 2024;56:2302980. doi: 10.1080/07853890.2024.2302980. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 5. Kumar VM, Keerthana A, Madhumitha M, et al. Sanative Chatbot For Health Seekers. IJECS . 2016;5:16022–5. doi: 10.18535/ijecs/v5i3.28. [ DOI ] [ Google Scholar ] 6. Schachner T, Keller R, V Wangenheim F. Artificial Intelligence-Based Conversational Agents for Chronic Conditions: Systematic Literature Review. J Med Internet Res . 2020;22:e20701. doi: 10.2196/20701. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 7. Sharma D, Kaushal S, Kumar H, et al. Chatbots in healthcare: challenges, technologies and applications. 2022 4th International Conference on Artificial Intelligence and Speech Technology (AIST); Delhi, India. pp. 1–6. [ DOI ] [ Google Scholar ] 8. Milne-Ives M, de Cock C, Lim E, et al. The Effectiveness of Artificial Intelligence Conversational Agents in Health Care: Systematic Review. J Med Internet Res . 2020;22:e20346. doi: 10.2196/20346. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 9. Laymouna M, Ma Y, Lessard D, et al. Roles, Users, Benefits, and Limitations of Chatbots in Health Care: Rapid Review. J Med Internet Res . 2024;26:e56930. doi: 10.2196/56930. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 10. Kolanu S, Dutta SJ, Roy S, et al. A Diabetic Diet Suggester and Appointment Scheduler Chatbot using Artificial Intelligence and Cloud. IRJASH . 2021;3:77–81. doi: 10.47392/irjash.2021.170. [ DOI ] [ Google Scholar ] 11. Goel R, Goswami RP, Totlani S, et al. Machine learning based healthcare chatbot. 2022 2nd International Conference on Advance Computing and Innovative Technologies in Engineering (ICACITE); Greater Noida, India. 2022. pp. 188–92. [ DOI ] [ Google Scholar ] 12. Haque A, Chowdhury M, Soliman H. Transforming chronic disease management with chatbots: key use cases for personalized and cost-effective care. 2023 Sixth International Symposium on Computer, Consumer and Control (IS3C); Taichung, Taiwan. pp. 1–6. [ DOI ] [ Google Scholar ] 13. Cheng G. ChatGPT: Principles and Architecture. Elsevier; 2025. [ Google Scholar ] 14. McGreevey JD, III, Hanson CW, III, Koppel R. Conversational Agents in Health Care—Reply. JAMA. 2020;324:2444. doi: 10.1001/jama.2020.21518. [ DOI ] [ PubMed ] [ Google Scholar ] 15. Kang J, Thompson RF, Aneja S, et al. National Cancer Institute Workshop on Artificial Intelligence in Radiation Oncology: Training the Next Generation. Pract Radiat Oncol. 2021;11:74–83. doi: 10.1016/j.prro.2020.06.001. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 16. Thirunavukarasu AJ, Ting DSJ, Elangovan K, et al. Large language models in medicine. Nat Med. 2023;29:1930–40. doi: 10.1038/s41591-023-02448-8. [ DOI ] [ PubMed ] [ Google Scholar ] 17. Chow JCL, Li K. Large Language Models in Medical Chatbots: Opportunities, Challenges, and the Need to Address AI Risks. Information . 2025;16:549. doi: 10.3390/info16070549. [ DOI ] [ Google Scholar ] 18. Chow JCL, Li K. Ethical Considerations in Human-Centered AI: Advancing Oncology Chatbots Through Large Language Models. JMIR Bioinform Biotech . 2024;5:e64406. doi: 10.2196/64406. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 19. Vaidyam AN, Wisniewski H, Halamka JD, et al. Chatbots and Conversational Agents in Mental Health: A Review of the Psychiatric Landscape. Can J Psychiatry. 2019;64:456–64. doi: 10.1177/0706743719828977. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 20. Vaidyam AN, Linggonegoro D, Torous J. Changes to the Psychiatric Chatbot Landscape: A Systematic Review of Conversational Agents in Serious Mental Illness: Changements du paysage psychiatrique des chatbots: une revue systématique des agents conversationnels dans la maladie mentale sérieuse. Can J Psychiatry. 2021;66:339–48. doi: 10.1177/0706743720966429. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 21. Geoghegan L, Scarborough A, Wormald JCR, et al. Automated conversational agents for post-intervention follow-up: a systematic review. BJS Open. 2021;5:zrab070. doi: 10.1093/bjsopen/zrab070. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 22. Bibault JE, Chaix B, Nectoux P, et al. Healthcare ex Machina: Are conversational agents ready for prime time in oncology? Clin Transl Radiat Oncol . 2019;16:55–9. doi: 10.1016/j.ctro.2019.04.002. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 23. Wang A, Qian Z, Briggs L, et al . IJGM 2023;Volume 16:1591–602. 10.2147/IJGM.S408208. [ DOI ] [ PMC free article ] [ PubMed ] 24. Xu L, Sanders L, Li K, et al. Chatbot for Health Care and Oncology Applications Using Artificial Intelligence and Machine Learning: Systematic Review. JMIR Cancer. 2021;7:e27850. doi: 10.2196/27850. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 25. Mills R, Mangone ER, Lesh N, et al. Chatbots to Improve Sexual and Reproductive Health: Realist Synthesis. J Med Internet Res. 2023;25:e46761. doi: 10.2196/46761. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 26. Balaji D, He L, Giani S, et al. Effectiveness and acceptability of conversational agents for sexual health promotion: a systematic review and meta-analysis. Sex Health. 2022;19:391–405. doi: 10.1071/SH22016. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 27. Altamimi I, Altamimi A, Alhumimidi AS, et al. Artificial Intelligence (AI) Chatbots in Medicine: A Supplement, Not a Substitute. Cureus. 2023;15:e40922. doi: 10.7759/cureus.40922. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 28. Abdekhoda M, Dehnad A. Applications of Chatbots in Improving Patient Care Outcomes: A scoping review. Sultan Qaboos Univ Med J. 2026;26:163–72. doi: 10.18295/2075-0528.2970. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 29. Haque MDR, Rubya S. An Overview of Chatbot-Based Mobile Mental Health Apps: Insights From App Description and User Reviews. JMIR Mhealth Uhealth. 2023;11:e44838. doi: 10.2196/44838. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 30. Ngo TTA, Phan TYN, Nguyen TK, et al. Understanding continuance intention toward the use of AI chatbots in customer service among generation Z in Vietnam. Acta Psychol (Amst) 2025;259:105468. doi: 10.1016/j.actpsy.2025.105468. [ DOI ] [ PubMed ] [ Google Scholar ] 31. Aiwerioghene EM, Osuchukwu VC. The Role of Artificial Intelligence in Healthcare Quality Improvement: A Scoping Review and Critical Appraisal of Operational Efficiency, Patient Outcomes, and Implementation Challenges. Hospitals. 2025;2:27. doi: 10.3390/hospitals2040027. [ DOI ] [ Google Scholar ] 32. Aytekin A, Alan H, Demirel H, et al. Digital Health Technologies in Patient Experience Literature: A Scoping Review and Future Outlook for Sustainable Digital Health Interventions. Sustainability. 2025;17:456. doi: 10.3390/su17020456. [ DOI ] [ Google Scholar ] 33. Santafé AS, Aranda JM, Beltrán WJ, et al. Technological Strategies for the Patient Experience in Emergency Departments: Scoping Review. JMIR Med Inform. 2026;14:e79782. doi: 10.2196/79782. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 34. Shamseer L, Moher D, Clarke M, et al. Preferred reporting items for systematic review and meta-analysis protocols (PRISMA-P) 2015: elaboration and explanation. BMJ. 2015;350:g7647. doi: 10.1136/bmj.g7647. [ DOI ] [ PubMed ] [ Google Scholar ] 35. McInerney P, Munn Z, Tricco AC, et al. Scoping Reviews. JBI Manual for Evidence Synthesis. 2020 doi: 10.46658/JBIMES-24-09. [ DOI ] [ PubMed ] [ Google Scholar ] 36. EndNote . Philadelphia (PA): Clarivate; [ Google Scholar ] 37. Rayyan Rayyan: AI-powered systematic review management platform. 2025. https://www.rayyan.ai Available. 38. van Dijk SHB, Brusse-Keizer MGJ, Bucsán CC, et al. Artificial intelligence in systematic reviews: promising when appropriately used. BMJ Open. 2023;13:e072254. doi: 10.1136/bmjopen-2023-072254. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 39. McKeown S, Mir ZM. Considerations for conducting systematic reviews: evaluating the performance of different methods for de-duplicating references. Syst Rev. 2021;10:38. doi: 10.1186/s13643-021-01583-y. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 40. ASReview LAB Developers ASReview LAB - A tool for AI-assisted systematic reviews (v2.0.1): Zenodo. 2025. [ DOI ] 41. van de Schoot R, de Bruin J, Schram R, et al. An open source machine learning framework for efficient and transparent systematic reviews. Nat Mach Intell . 2021;3:125–33. doi: 10.1038/s42256-020-00287-7. [ DOI ] [ Google Scholar ] 42. Settles B. University of Wisconsin–Madison; 2009 Technical Report. 2009. https://burrsettles.com/pub/settles.activelearning.pdf Available. 43. Boetje J, van de Schoot R. The SAFE procedure: a practical stopping heuristic for active learning-based screening in systematic reviews and meta-analyses. Syst Rev. 2024;13:81. doi: 10.1186/s13643-024-02502-7. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 44. Bruin J, Ferdinands G, Harkema A, et al. ASReview software documentation. doi: 10.5281/zenodo.4287119. [ DOI ] 45. Teijema JJ, Hofstee L, Brouwer M, et al. Active learning-based systematic reviewing using switching classification models: the case of the onset, maintenance, and relapse of depressive disorders. Front Res Metr Anal . 2023;8:1178181. doi: 10.3389/frma.2023.1178181. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 46. Elicit Elicit: The AI Research Assistant. https://elicit.com Available. 47. Bui QH, Trinh TA, Le Thi HA, et al. Towards a sustainable university transition model for emerging markets. Cogent Business & Management. 2023;10:2272372. doi: 10.1080/23311975.2023.2272372. [ DOI ] [ Google Scholar ] 48. Kung JY. Elicit. J Can Health Libr Assoc. 2023;44:15–8. doi: 10.29173/jchla29657. [ DOI ] [ Google Scholar ] 49. Clinical Outcome - Toolkit National Center for Advancing Translational Sciences Toolkit for Patient Focused Therapy Development. 2020. https://toolkit.ncats.nih.gov/glossary/clinical-outcome Available. 50. Yahagi M, Hiruta R, Miyauchi C, et al. Comparison of Conventional Anesthesia Nurse Education and an Artificial Intelligence Chatbot (ChatGPT) Intervention on Preoperative Anxiety: A Randomized Controlled Trial. J Perianesth Nurs. 2024;39:767–71. doi: 10.1016/j.jopan.2023.12.005. [ DOI ] [ PubMed ] [ Google Scholar ] 51. Gomes CF, Yasin MM, Yasin Y. Assessing operational effectiveness in healthcare organizations: a systematic approach. Int J Health Care Qual Assur. 2010;23:127–40. doi: 10.1108/09526861011017067. [ DOI ] [ PubMed ] [ Google Scholar ] 52. National Institutes of Health Data Synthesis: National Institutes of Health. 2025. https://www.nihlibrary.nih.gov/services/evidence-synthesis/select-collect-synthesize/data_synthesis Available. 53. Ivanov S, Webster C. In: Robots, artificial intelligence, and service automation in travel, tourism and hospitality. Ivanov S, Webster C, editors. Emerald Publishing Limited; 2019. Robots, artificial intelligence, and service automation in travel, tourism and hospitality ; pp. 39–55. https://www.emerald.com/insight/publication/doi/10.1108/9781787566873 Available. [ Google Scholar ] 54. Bickmore TW, Kimani E, Trinh H, et al. Managing chronic conditions with a smartphone-based conversational virtual agent. IVA ’18; Sydney NSW Australia. Nov 5, 2018. pp. 119–24. Available. [ DOI ] [ Google Scholar ] 55. Pereira J, Díaz Ó. Using Health Chatbots for Behavior Change: A Mapping Study. J Med Syst. 2019;43:135. doi: 10.1007/s10916-019-1237-1. [ DOI ] [ PubMed ] [ Google Scholar ] 56. Chan Y-T, Abad JE, Dibart S, et al. Assessing the article screening efficiency of artificial intelligence for Systematic Reviews. J Dent. 2024;149:105259. doi: 10.1016/j.jdent.2024.105259. [ DOI ] [ PubMed ] [ Google Scholar ] 57. Waffenschmidt S, Knelangen M, Sieben W, et al. Single screening versus conventional double screening for study selection in systematic reviews: a methodological systematic review. BMC Med Res Methodol. 2019;19:132. doi: 10.1186/s12874-019-0782-0. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 58. O’Mara-Eves A, Thomas J, McNaught J, et al. Using text mining for study identification in systematic reviews: a systematic review of current approaches. Syst Rev. 2015;4:5. doi: 10.1186/2046-4053-4-5. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 59. Ruan M, Fan J, Liu M, et al. Artificial intelligence for the science of evidence synthesis: how good are AI-powered tools for automatic literature screening? BMC Med Res Methodol. 2025;25:199. doi: 10.1186/s12874-025-02644-9. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] BMJ Open. 2026 Jun 3. Review Process File Copyright and License information PMC Copyright notice bmjopen-2025-113547.reviewer_comments.pdf (351KB, pdf) Open in a new tab Associated Data This section collects any data citations, data availability statements, or supplementary materials included in this article. Supplementary Materials online supplemental file 1 bmjopen-16-6-s001.pdf (226KB, pdf) DOI: 10.1136/bmjopen-2025-113547 Articles from BMJ Open are provided here courtesy of BMJ Publishing Group ACTIONS View on publisher site PDF (348.0 KB) Cite Collections Permalink RESOURCES Similar articles Cited by other articles Links to NCBI Databases Follow NCBI NCBI on X (formerly known as Twitter) NCBI on Facebook NCBI on LinkedIn NCBI on GitHub NCBI RSS feed Connect with NLM NLM on X (formerly known as Twitter) NLM on Facebook NLM on YouTube National Library of Medicine 8600 Rockville Pike Bethesda, MD 20894 Web Policies FOIA HHS Vulnerability Disclosure Help Accessibility Careers NLM NIH HHS USA.gov Back to Top

Related documents

• Evaluating the ability of AI chatbots to provide informed consent information for common oncological surgeries.#331270
• Enhancing patient care with AI agents: integrating advanced chatbot technologies for improved healthcare delivery.#1175516
• Large language model chatbots as sources of pediatric anesthesia health advice: An evaluation of reliability and readability.#334833
• Reduced symptom reporting quality during human-chatbot versus human-physician interactions#167753
• Trust and verification in AI-enabled physician chatbots for chronic disease management: evidence from digital health behavior.#354661