Generative artificial intelligence (GenAI), a subset of artificial intelligence (AI), leverages large language models (LLMs) to learn patterns from vast datasets and generate novel content [1-3].

In health care, the use of GenAI tools for clinical documentation is growing, primarily scribes and chatbots. Examples of ambient AI scribes, also known as ambient voice technology, in the market include Nuance Dragon Ambient eXperience by Microsoft [4], Speke by ScribeAmerica [5] and Tandem Health [6], among others.

Chatbots range from general purpose models that are not clinically validated, such as OpenAI’s ChatGPT, to medical-specific ones, like Med-Palm, developed by Google Research [7], which is pretrained on medical data and has undergone clinical benchmarking in clinical settings.

Applied to health care, these tools may help clinicians with documentation tasks and reduce administrative burdens, which are known drivers of burnout [8-10]. These tools may also improve accessibility for patients [11,12] by synthesizing complex data and writing responses in a desired conversational style or literacy level [13,14], ultimately improving the overall efficiency of health care services [13,15].

The adoption of GenAI in health care, however, also raises significant concerns [15,16]. AI outputs can be biased, potentially exacerbating ageism, racism, or gender discrimination in health care due to omissions and biases present in the training data used [17-20]. Such “algorithmic biases” can result in unfair suggestions and discriminatory advice. Even in passive use cases, such as ambient scribes that transcribe clinician-patient interactions, bias may emerge in how speech is transcribed, summarized, or prioritized, potentially misrepresenting certain dialects, accents, or phrasing more commonly used by specific demographic groups. This could affect the completeness or accuracy of documentation and introduce disparities into medical records. These tools may also generate dangerous medical advice and can produce incorrect information called “hallucinations” [20,21]. Users may be vulnerable to misinformation because of the compelling, rapid responses offered, which could further endanger privacy and safety [16,22].

Previous reviews have examined the perspectives of health care professionals (HCPs) on AI use in clinical settings, focusing on its roles in decision-making, efficiency, quality of care, and documentation systems [23,24]. Others have explored perceived benefits and concerns, such as bias, transparency, data privacy, and safety, as well as factors influencing AI adoption, including barriers and facilitators influencing AI adoption [25-28]. Additional review studies have investigated HCPs’ overall understanding and experiences with AI, mapping the broader implications of emerging AI-powered technologies across the health care landscape [29,30]. However, none of these reviews have specifically examined the lived experiences, perceptions, or opinions of HCPs, particularly frontline clinicians, regarding the use of GenAI in clinical documentation and decision-making.

Added to this gap, survey research indicates important trends and insights in relation to the adoption of technological tools for clinical documentation. Namely, HCPs both desire and expect technological tools to alleviate their administrative burdens [8-10].

Theoretical framing is also required to better explore barriers and facilitators to GenAI uptake in clinical documentation. For example, the technology acceptance model (TAM) provides one useful framework to examine this gap, highlighting perceived usefulness and ease of use as key factors influencing technology adoption [31].

To address this theoretical and empirical gap, we have planned a scoping review to systematically map the current evidence on HCPs’ experiences and opinions of GenAI use in clinical documentation. This manuscript presents the protocol detailing the methodology and approach that will guide the review.

This scoping review aims to explore HCPs’ perspectives on the use of GenAI tools, such as chatbots and ambient AI, specifically for clinical documentation, by synthesizing existing evidence on their experiences and opinions and identifying gaps to guide future research.

To achieve this aim, the following objectives have been developed: (1) to identify, collate, and summarize existing evidence on HCPs’ experiences of GenAI tools, including chatbots and ambient AI, on clinical documentation; (2) to assess HCPs’ opinions about, and perceived impact of, GenAI on documentation efficiency, quality, privacy, and the potential harms and benefits for HCPs and patients; and (3) to identify research gaps to inform future investigations into HCPs’ experiences and opinions on the use of GenAI and ambient AI for clinical documentation.

A scoping review is well-suited for exploring emerging fields like GenAI in clinical documentation, as it maps the existing literature, identifies key themes, and highlights gaps in knowledge. Unlike a systematic review, which focuses on a narrowly defined question with strict inclusion criteria, a scoping review provides a broader overview, capturing diverse perspectives, methodologies, and contexts [32]. This approach is valuable for informing future research directions and policy considerations in a rapidly evolving technological landscape [32]. We will carry out a scoping review using the framework proposed by Arksey and O’Malley [32], consisting of the following five stages: (1) identifying the research question, (2) identifying the relevant studies, (3) selecting eligible studies, (4) collecting data, and (5) summarizing data and synthesizing results. The Preferred Reporting Items for Systematic Reviews and Meta-Analyses Extension for Scoping Reviews (PRISMA ScR) checklist will be used to report the review [33,34]. The final publication will emphasize any further changes made to the study design.

Based on the objectives and iterative discussions within the research team, we agreed on the primary research question (RQ) and 8 secondary RQs.

The PRISMA flow diagram [34] outlines the planned multistage screening process for identifying relevant studies (Figure 1). As this is a protocol, the flowchart includes placeholders marked “n=0” to indicate that specific data (eg, number of records screened or excluded) are not yet available, since the formal search and selection process has not been conducted. An experienced research librarian from Uppsala University, Sweden, will handle the deduplication procedure.

The search strategy was developed through an iterative process by CGS and the Uppsala University library. The first list of search phrases was compiled using seed references, and it was further refined through conversations between the library and CGS. It is composed of four main concepts: (1) HCPs, (2) GenAI, (3) clinical documentation, and (4) experiences and opinions. These concepts were combined with the Boolean AND and OR operators, as shown in Textbox 2. The literature search will be done in the following five databases: PubMed, IEEE Xplore, APA PsycInfo, CINAHL, and Web of Science. The search terms were tailored to different databases. The whole search query is available in Multimedia Appendix 1.

After selecting the studies to include, metadata (eg, title, authors, and publication year) of the remaining records will be exported and summarized to Google Sheets for further processing. The spreadsheet will be extended with the following parameters, based on the studies’ full texts: publication type, country, sample characteristics, clinical setting, study aim, study design, research question, and key findings. This will allow us to extract and organize relevant data from included studies. The results will be presented in diagrams or tabular format for a clear and comprehensive visualization. While no design restrictions will be applied during the search phase to ensure broad retrieval of relevant studies, study design will be evaluated during the screening and data charting processes. To manage potential heterogeneity, included studies will be grouped according to their methodological design (eg, qualitative, quantitative, or mixed methods) and analyzed within these categories during the synthesis phase. This approach allows us to preserve the comprehensiveness of the search while maintaining analytical clarity and relevance in reporting.

All members of the study team will participate in the data extraction process. Additionally, the first author (CGS) will verify that the data extraction is accurate and comprehensive. The Mixed Methods Appraisal Tool (MMAT) [39] will be used to evaluate the studies’ quality and methodological rigor. All studies included will be graded by two researchers separately, and their findings will be approved by a third independent researcher who will be consulted if no consensus can be obtained.

This study is not subject to ethical approval since only publicly available data material will be used within the scoping review methodology.

This scoping review is expected to provide a comprehensive overview of the current evidence on HCPs’ experiences and opinions regarding the use of GenAI tools in clinical documentation. By systematically mapping this literature, we aim to identify prevailing themes as well as significant knowledge gaps. For example, we anticipate insights into perceived risks related to patient safety, bias, and data protection, as well as ethical and privacy concerns that may influence technology acceptance and use in clinical settings.

By applying the TAM as a guiding framework, we will interpret patterns across studies related to perceived usefulness, ease of use, and behavioral intention to adopt GenAI tools. This approach may reveal aspects such as usability challenges or contextual factors shaping HCPs’ readiness to integrate these technologies.

Furthermore, our review could help clarify how ethical considerations overlap with user acceptance, practical implementation, and broader sociotechnical dynamics that influence GenAI uptake in health care. This integrated perspective can inform targeted strategies to support safe, ethical, and effective adoption of GenAI tools in clinical practice.

Finally, applying the MMAT to assess study quality will support a more robust and trustworthy narrative synthesis by contextualizing findings within the methodological strength of the available evidence.

Several reviews have explored HCPs’ perspectives on AI in clinical settings. Ayorinde et al [23] conducted a systematic review synthesizing experiences with non–knowledge-based AI tools in decision-making. Bracken et al [24] focused on the efficiency, quality, and stakeholder views of AI-driven documentation systems using measurable outcomes. Buchanan et al [29] mapped emerging AI-powered health technologies and their implications for nursing across clinical, administrative, policy, and research domains. Chustecki [25] used a narrative review to examine benefits and concerns related to AI in health care, including bias, transparency, data privacy, and safety. Hassan et al [26] and Lambert et al [27] conducted scoping and integrative reviews, respectively, to identify barriers and facilitators to AI adoption and acceptance among HCPs. Sahoo et al [28] highlighted perceived pros and cons of AI in health care, while Shinners et al [30] explored HCPs’ understanding of and experiences with AI in care delivery. While these reviews broadly explore HCPs’ perspectives on AI, none specifically address their experiences and opinions on the use of GenAI for clinical documentation.

Given that GenAI tools are increasingly being heralded as valuable in clinical documentation, this scoping review examines an urgent topic. By synthesizing HCPs’ experiences and opinions, it strives to obtain insights on implementation and uptake of these tools. These findings can guide future research and support the safe integration of GenAI in health care.

As GenAI rapidly evolves, a potential limitation of the study is that the findings may soon become outdated. Additionally, the exclusion of non–English-language studies might lead to the omission of relevant research. This may limit the generalizability of the findings and introduce a Western-centric bias, particularly given the global scale of GenAI adoption in health care. This constraint may overlook relevant insights from non–English-speaking countries. While resource limitations prevent full inclusion of non–English-language literature, studies with English-language abstracts will be screened when available to partially mitigate this limitation. Furthermore, the exclusion of gray literature means that some studies will be overlooked. Nevertheless, our search strategy is designed to identify the most comprehensive and high-quality evidence.

Our scoping review aims to map current research on this issue. This approach aims to clarify how HCPs are—or are not—integrating GenAI into their daily work, helping to identify perceived benefits, risks, expectations, and uncertainties. Results also may help to identify research gaps in this field.

To ensure the findings of this scoping review reach relevant stakeholders and contribute meaningfully to practice, policy, and future research, we plan to implement a comprehensive dissemination strategy. This will include publishing the results in a peer-reviewed, open-access journal and presenting them at relevant academic conferences and professional meetings.

This scoping review will systematically map the existing literature on HCPs’ experiences and opinions regarding the use of GenAI tools in clinical documentation. By identifying key themes, knowledge gaps, and patterns related to technology acceptance, the findings will help inform future research, policy, and practice. The integration of the TAM as a guiding framework, along with focused attention on ethical and practical considerations, will provide a structured lens for interpreting the literature. Ultimately, this review aims to support the safe, effective, and ethical implementation of GenAI in health care settings.