Artificial intelligence (AI) refers to software systems capable of performing tasks that mimic human reasoning through data processing and algorithms (Kaplan et al., 2023; West, 2018). As open-source AI platforms have expanded, adoption across sectors—including healthcare—has accelerated, heightening the need to understand and establish trust in these technologies (Lyons et al., 2024). When appropriately applied, AI can enhance task accuracy, improve safety, and support decision-making efficiency (Baduge et al., 2022; Sardar et al., 2019; Soori et al., 2023; Abdar et al., 2022; Begoli et al., 2019). In healthcare, AI is increasingly used to streamline workflows and reduce human error. Examples include electronic medical record (EMR) automation to support coordinated care (Rojahn et al., 2023), clinical decision support systems (CDSS) that aid diagnosis and treatment planning (Laxar et al., 2023; Osheroff et al., 2007), and robot-assisted surgery designed to minimize procedural risks (Hamet and Tremblay, 2017; Soori et al., 2023). Despite these benefits, challenges remain, such as inaccurate chatbot responses, diagnostic errors, and risks of exacerbating health inequities (Kaplan et al., 2023; D’Elia et al., 2022). Growing concerns around reliability, bias, transparency, and ethical use make trust a critical component of successful AI integration (Kostick-Quenet et al., 2024; Asan et al., 2020).

Trust involves vulnerability and uncertainty and shapes how individuals engage with automated systems as much as with other people (Deng et al., 2024; Hoff and Bashir, 2015). Both insufficient trust—leading to system avoidance—and overtrust—leading to misuse—pose risks (Muir and Moray, 1996). In healthcare, trust underpins patient-provider relationships and clinician reliance on diagnostic and treatment supports (Birkhäuer et al., 2017; Gaube et al., 2021). Trust has long been fundamental to healthcare, shaping both patient–clinician relationships and clinical decision-making processes. Patients must trust that clinicians act in their best interests and possess the expertise to deliver safe, effective care (Birkhäuer et al., 2017). Likewise, clinicians must have confidence in the tools and systems that support their diagnostic and treatment decisions (Gaube et al., 2021). As AI becomes increasingly embedded in clinical workflows, establishing trust from both patients and clinicians is essential for adoption and meaningful use (Asan et al., 2020). Ultimately, trust determines whether AI technologies are accepted, integrated into practice, and relied upon in patient care, underscoring the need to understand the evolving dynamics of human-machine trust in healthcare (Asan et al., 2020; Gaube et al., 2021; Lee and See, 2004). Therefore, fostering trust in AI requires addressing the needs and expectations of both clinicians and patients (Asan et al., 2020).

This rapid review synthesizes current evidence on the factors shaping human-machine trust in healthcare, with the aim of informing safe, effective, and trusted AI implementation.

The search strategy for this rapid review was developed in consultation with a Health Sciences Librarian. Searches were conducted on March 22, 2024, in MEDLINE (Ovid), Embase (Ovid), and CINAHL (Ebsco). These databases were chosen for this rapid review due to their extensive coverage and relevance to medical and allied health literature, especially in relation to the application of artificial intelligence in healthcare.

To capture the scope of the published research, the final search strategy relies on Medical Subject Headings (MeSH) from MEDLINE (Ovid) and is translated to the equivalent term in the other resources. Target articles were also reviewed for relevant subject headings. Previous iterations of the search strategy were complex, and many citations were found to be too specific for this review. Because of the complex and individualized nature of trust relationships, we decided not to define trust before commencing our study (Table 1).

AI holds significant potential to enhance decision-making, increase efficiency, and transform existing healthcare practices. Research indicates that the impact of automated systems will depend on whether and how human factors are incorporated into their design (Asan et al., 2020). Thus, as AI continues to be implemented throughout the healthcare system, considering the factors that impact patient and physician trust in AI technologies is of vital importance (Baduge et al., 2022; Glikson and Woolley, 2020). Trust is a complex construct involving many factors, and it plays a critical role in both human relationships and human-machine interactions (Asan et al., 2020). Our review identified that only 35% of the included studies provided an explicit definition of trust. Furthermore, none of the articles utilized the same definition, underscoring the complexity of evaluating this variable using a standardized approach. Definitions of trust were subsequently categorized according to the context under evaluation, including trust in technology, psychological perspectives on trust, and trust within healthcare settings. Across various disciplines, workers’ trust in AI technology is an important aspect of the successful integration of AI into a workplace (Glikson and Woolley, 2020). This rapid review addresses the factors that contribute to developing a trusting relationship in AI technologies within the healthcare industry.

Several factors were identified as essential prerequisites for fostering trust in AI technologies, highlighting the minimum requirements that AI systems must meet to gain user confidence. These factors include AI technologies addressing privacy concerns, maintaining accuracy, providing credible information, high levels of system performance, addressing biases, and overall reliability (Bergquist et al., 2024; Fischer et al., 2023; Kostick-Quenet et al., 2024; Stevens and Stetson, 2023).

When users understood the capabilities and limitations of the AI system they were using, a trusting relationship could be developed (Louca et al., 2023; Moilanen et al., 2023; Viberg Johansson et al., 2024). Both patients and physicians expressed a desire to understand how the overall system was developed and how datasets are utilized to build and refine the algorithms used in healthcare settings (Asokan et al., 2023; Bergquist et al., 2024; Kostick-Quenet et al., 2024; Liu et al., 2022). Gender was linked to attitudes toward AI technology, with females having less trust and more negative attitudes toward technology compared to males in all but one study (Hsieh, 2023) where gender was insignificant in predicting physician trust (Çitil and Çitil Canbay, 2022; Hsieh, 2023; Kim and Kim, 2022; Liu and Tao, 2022; Sonmez, 2024). Sonmez (2024) specifically found that women had a higher risk aversion compared to men, with less positive attitudes, regardless of the human-likeness of AI. Additionally, experience using technology was a contributing factor leading to trust in AI technology. Those with more experience with technology were more likely to trust AI technology and their decisions (Asokan et al., 2023; Moilanen et al., 2023; Schulz et al., 2023). In contrast, previous negative experiences caused distrust and deterred the use of the systems (Nash et al., 2023). Although some studies report the impact of age on trust in technology, one study in this review found that neither prior experience nor age was a factor impacting trust in AI systems or intentions to utilize AI systems (Choudhury and Asan, 2023). This contradicts findings that demonstrated a significant impact of age on trust in AI technology, with older individuals generally exhibiting lower levels of trust in AI compared to younger individuals (Nash et al., 2023; Oksanen et al., 2020). Overall, the more familiar a user is with AI technology, the more likely they are to trust the system. Many studies have found explainable artificial intelligence to be an essential factor impacting user trust (Bergquist et al., 2024; Fischer et al., 2023; Goel et al., 2022; Helenason et al., 2024; Qassim et al., 2023). ^As outlined in Theme 1, several included studies reported higher adoption and comfort when AI provided interpretable outputs rather than “black box” predictions. This underscores explainability as a critical design feature for trustworthy AI in healthcare (Bergquist et al., 2024; Fischer et al., 2023; Goel et al., 2022; Helenason et al., 2024; Qassim et al., 2023). However, two studies found that explainability was not a factor that predicted whether physicians would or would not utilize AI technology and did not act as a factor to improve trust in healthcare contexts (Lancaster Farrell, 2022; Liu et al., 2022). Understanding the complexity of trust and how trust is built was another important factor influencing trust relationships. For example, trust relationships were impacted when users of AI technologies recognized that trust is a learned behavior and evaluated the overall impact of whether they were to trust or not trust a system (Hallowell et al., 2022).

Our review found that trust is built when the user believes there will be positive benefits from using the AI system. Positive perceptions regarding the usefulness of AI technologies resulted in greater user trust (Alhur et al., 2023; Stevens and Stetson, 2023). Specifically, physicians were likely to use and trust AI technologies if they found them to be efficient, easy to use, and useful in their practice (Asokan et al., 2023; Cheng et al., 2022; Hsieh, 2023; Liu and Tao, 2022; Nash et al., 2023; Qassim et al., 2023; Schulz et al., 2023; Shamszare and Choudhury, 2023; Yakar et al., 2022; Yi-No Kang et al., 2023). As the perceived value of an AI technology increases, physicians are more likely to rely on the system, thereby enhancing their trust in it (Bergquist et al., 2024; Goel et al., 2022; Liu et al., 2022). Furthermore, trust in the AI system was enhanced when the system could be seamlessly integrated into the physicians’ practice and demonstrated compatibility with existing systems (Bergquist et al., 2024).

The human-likeness of AI systems plays a key role in shaping patient trust, fostering a greater openness to incorporating AI technology into their care networks (Choudhury and Shamszare, 2023; Kim and Kim, 2022; Liu and Tao, 2022; Sonmez, 2024). The greater the degree of human-like attributes exhibited by AI technologies, such as surgical robots, the higher the likelihood that individuals will extend their trust in these systems, owing to an enhanced sense of social presence (Kim and Kim, 2022; Liu and Tao, 2022). Trust repair after system failure remains challenging for AI, though findings were mixed, as outlined in Theme 2 (Asokan et al., 2023; Moilanen et al., 2023). This reflects the psychological aspect of trust: human errors are often seen as part of a learning process, while AI mistakes are viewed as indicative of systemic flaws (Moilanen et al., 2023). Consequently, this underscores the necessity for effective trust repair strategies to rebuild confidence in AI systems within healthcare settings (Moilanen et al., 2023). This review found that different psychological and external influences can shape the feelings of the patient toward AI. Therefore, understanding these factors can help build patient acceptance of AI systems and build patient trust in AI even when it makes mistakes (Huo et al., 2022).

As detailed in Theme 2, trust in AI often follows a cascade: patients are more likely to trust systems endorsed by their physicians, and physicians are more comfortable when the technology has been validated by peers, experts, or reputable organizations. Rather than revisiting individual study findings, this reinforces the importance of human validation as a bridge between developers, clinicians, and patients (Asokan et al., 2023; Cheng et al., 2022; Fischer et al., 2023; Helenason et al., 2024; Hallowell et al., 2022; Kostick-Quenet et al., 2024; Qassim et al., 2023; Rodler et al., 2024; Sonmez, 2024). Trust was further enhanced if professionals outside the healthcare field, such as software engineers, program designers, and technology leaders, confirmed the trustworthiness of AI systems (Szabo et al., 2024). Physician trust would fluctuate, though, based on the result of the AI technology. For example, if the system results in decisions that the physician disagrees with, they may grow to distrust it (Hallowell et al., 2022). However, trust was enhanced if the system tended to come to the same decision as the physician (Asokan et al., 2023; Hallowell et al., 2022; Helenason et al., 2024; Qassim et al., 2023; Viberg Johansson et al., 2024). Additionally, physicians who recognize the potential for bias in the development of AI technologies tend to be more cautious and less likely to trust these systems sufficiently to integrate them into their clinical practice (Van Bulck and Moons, 2024). To trust a system, physicians want to be confident that the system is trained on a robust dataset and that these systems are constantly updating. This can lead to a disconnect between patients’ privacy of their data and accurately training the AI systems to be as accurate as possible (Bergquist et al., 2024). In a study conducted by Viberg Johansson et al. (2024) looking at women’s perceptions and attitudes toward the implementation of AI in mammography, patients were willing to share their data for life-saving measures or research purposes but were reluctant to share with private entities, as they considered it “an intrusion into their private lives.” Patients may not understand that data sharing with private entities is essential for technology advancements (Viberg Johansson et al., 2024).

The Organization for Economic Co-Operation and Development (OECD) states that trustworthy AI will be transparent and explainable, where the capabilities and limitations of the system are readily available and easily explained to users (OECD Legal Instruments, 2019). A major area of concern with AI in sensitive fields such as healthcare is the “black box” nature of the system. Output and decisions are often given without explanation behind them, making implementation and trust difficult (Liu et al., 2022). The introduction of explainable AI (XAI) refers to the actions and measures taken to ensure transparency in the AI system that is both explainable and interpretable (Adadi and Berrada, 2018; Liu et al., 2022). Implementing XAI could be a major step toward generating trust in AI systems and furthering their implementation. Trustworthy AI should have security and safety measures in place to avoid harm in the case of misuse or adverse conditions (OECD Legal Instruments, 2019). Finally, trustworthy AI should be accountable and ensure traceability of datasets, processes and decisions made by the system (OECD Legal Instruments, 2019). Because AI can be implemented in vastly different ways in different fields, major factors impacting trust in one field may vary greatly from those in another field. Understanding the key aspects that play a role in determining user trust in AI systems is important for the successful implementation of AI (Abeywickrama et al., 2023; Louca et al., 2023). The ability of the AI system to explain its decision-making process can reveal bias that a diagnosing practitioner may have been unaware of (Bergquist et al., 2024). Therefore, collaboration between technology programmers and healthcare professionals may ensure that the outputs of AI technology are valid and aligned with clinicians, resulting in stronger trust relationships (Nash et al., 2023). Furthermore, when patients believed the physician was making the final decision, not the AI system, they were more open to trusting AI technologies to assist with healthcare decisions (Rodler et al., 2024; Van Bulck and Moons, 2024; Viberg Johansson et al., 2024). However, studies found that patients preferred physicians over AI systems, even if trust was formed (Rojahn et al., 2023; Viberg Johansson et al., 2024). Thus, a cascading relationship exists in user trust for AI systems. For a patient to trust the system, the physician must first trust the system, and the physician’s trust depends on their confidence in the program developer who created it (Hallowell et al., 2022). From the patient’s perspective, a common barrier to trust in AI technologies is the concern over privacy invasion and data collection practices (Liu and Tao, 2022; Moilanen et al., 2023; Nash et al., 2023; Viberg Johansson et al., 2024). In contexts where users exhibit a moderate level of trust in AI systems, the implementation of direct safeguards for data protection and privacy significantly enhanced trust in these systems (Alhur et al., 2023). Healthcare organizations need to adopt standardized procedures for addressing data collection and management issues associated with AI technologies to foster trust among patients (Bergquist et al., 2024; Costantino et al., 2022). This includes physicians ensuring open communication and transparency with patients surrounding privacy and data security (Shahzad Khan et al., 2024).

To trust a system, physicians want to be confident that the system is trained on a robust dataset and that these systems are constantly updating. This can lead to a disconnect between patients privacy of their data and accurately training the AI systems to be as accurate as possible (Bergquist et al., 2024). Physicians may be more willing to share data for research purposes; however, patients may not understand that data sharing with private entities is essential for technology advancements (Viberg Johansson et al., 2024). Overall, the successful integration of AI technologies into healthcare will depend on fostering trust among both physicians and patients. The multifaceted nature of trust is evident, as well as the importance of transparency, system reliability, and privacy protections. As the development and implementation of AI evolves, understanding and addressing the factors that influence trust will be critical to ensure effective adoption and long-term success in improving healthcare outcomes.

This study has several limitations. Firstly, as this study is a rapid review, the search was less comprehensive than a scoping or systematic review. The scope of this study was limited to healthcare contexts, so we caution against any interpretations of findings beyond the healthcare setting. Future studies should take a broader approach to consider factors of trust in other contexts. Our search was limited to academic publications, so there could be valuable information in the gray literature that was excluded from this review.

This review does not address methods for building trust between humans and artificial intelligence in healthcare settings; thus, future research should address this. Further studies should examine the factors of trust in AI outside of healthcare settings. Additionally, longitudinal studies should examine how trust evolves and explore effective recovery strategies related to AI errors. These studies could further investigate how initial impressions of AI evolve with sustained use. This may include investigating effective trust repair mechanisms, including system transparency, error communication strategies, and human oversight models, to restore confidence after AI failure. Future studies may adopt large-language models (LLMs) in order to assist with literature analysis. For the purposes of this review LLMs were not utilized as our goal was to ensure transparency, reproducibility, and methodological rigor by relying on traditional, human-driven review methods such as systematic searching, critical appraisal, and manual synthesis of findings. While LLMs have shown promise for assisting in scientific research and paper drafting, their capabilities, limitations, and best practices for reliable use in scholarly reviews were not yet well defined at the time of this work, and integrating them was considered outside the scope of our study.

An effort should be made to develop standardized metrics and validated scales to measure trust in AI systems consistently across healthcare contexts. The current literature varies and often involves non-comparable definitions of trust, making cross-study synthesis challenging. It also suggested that a co-design and participatory development framework is utilized, engaging clinicians, patients, and technology developers to ensure that AI tools align with user needs, workflows, and ethical expectations. Future studies should also investigate how trust is shaped across diverse cultural contexts and low-resource settings, as well as how trust develops within interdisciplinary teams that include not only clinicians but also information technology developers, administrators, and other stakeholders involved in AI implementation. Although this rapid review did not aim to evaluate or propose specific methodological frameworks, future work would benefit from the development and application of structured approaches—such as standardized trust metrics, longitudinal study designs to track trust over time, and co-design models that engage clinicians, patients, and developers—to guide the creation and assessment of trustworthy AI systems.

Through a rapid review of the literature, our findings suggest that trust is imperative for the successful implementation of AI in healthcare settings and that there are numerous factors that contribute to patient and physician trust in AI systems. By examining broad categories that shape trust, including AI literacy, psychology and utility, this review underscores the complexity of building trust across diverse user groups. Trust in AI is not a static construct that is well defined but involves a dynamic interplay of system transparency, perceived value, and user experience. Furthermore, concepts such as anthropomorphism, privacy, and trust repair strategies must be considered when evaluating trust in AI systems. In order to use AI systems to their full potential in healthcare settings, developers, clinicians, and policymakers must collaborate to address these multifaceted aspects of trust. This involves designing AI systems that are transparent, explainable, and well-aligned with ethical standards to maintain robust privacy safeguards. Future research should continue exploring the nuanced relationships among trust factors and expand beyond healthcare to inform AI applications in other critical domains. Trustworthy AI systems not only have the ability to enhance clinical decision-making and efficiency but could pave the way for a more equitable and inclusive healthcare landscape.