Integrating artificial intelligence (AI) in healthcare: advancing older adults’ health management in Saudi Arabia through AI-powered chatbots

AI-powered chatbots in healthcare

AI-powered chatbots provide versatile features, including real-time assistance, tailored information, and uninterrupted connectivity. Chatbots enhance patient engagement by providing appointment reminders, integrating electronic medical records (EMRs) for customized health information, and offering guidance on healthy lifestyle choices, including diet, exercise, and illness prevention. These applications collect data from diverse sources, assisting patients in managing their medical conditions with personalized guidance and timely reminders about medications and physical activities (Aggarwal et al., 2023). Moreover, these chatbots can seamlessly integrate with virtual and augmented reality platforms, delivering immersive experiences that can accelerate and enhance health behavior modifications (Aggarwal et al., 2023). Figure 1 illustrates examples of scenarios for using chatbots in healthcare, with diverse situations enabling the transformation of healthcare services.

The global tendency of adopting AI chatbots in healthcare reflects a growing trend across different types of healthcare systems. In the United States, AI-driven virtual health assistants have been integrated into chronic disease management programs to help with automated monitoring and behavioral coaching of patients (Igwama et al., 2024; Pan et al., 2025). In Japan, socially assistive AI chatbots and other solutions are widely considered for implementation in eldercare facilities to improve patient well-being through interactive companionship and cognitive stimulation (Ho et al., 2023). While these solutions address both medical and emotional needs, low-resource settings are yet to effectively embrace AI technology for primary healthcare support. Nevertheless, innovative approaches are emerging to bridge this gap. AI-driven data analytics and predictive modeling hold promise in bridging these gaps by optimizing resource allocation, enabling early disease detection, and tailoring interventions to specific geriatric populations, even in infrastructure-limited environments. (Odionu & Ibeh, 2024; Zhao & Li, 2024).

Digital transformation and the digital divide

Digital transformation in healthcare, integrating technologies like telemedicine, AI, virtual consultations, and the Internet of Things (IoT), aims to sustain access, improve service delivery, and enhance patient engagement (Mulukuntla & Venkata, 2020). However, this shift has exacerbated the digital divide among older adults, making them vulnerable due to inadequate digital literacy and limited access to reliable technologies. Defined as the disparity in access to digital technologies influenced by socioeconomic status, geographic location, and age (Jaarsveld, 2020), the digital divide poses significant challenges for the elderly, particularly in healthcare. Despite advancements in AI applications that enhance independence and encourage participation in care, digital exclusion remains a critical issue. Older adults face substantial barriers in using new technologies, impacting their health outcomes as more services move online (Carrasco-Dajer et al., 2024; Chu et al., 2022; Sriwisathiyakun & Dhamanitayakul, 2022). Addressing these barriers is crucial for fostering inclusion and enhancing the quality of life, allowing them to live independently at home longer.

A significant issue elderly people encounter when utilizing AI and chatbots is a lack of digital literacy. Pearce (2024) asserts that the digital literacy gap may result in dissatisfaction and disengagement, hindering older individuals from accessing essential health information or services that could enhance their well-being. Additionally, research indicates that the complexity of numerous AI-driven platforms frequently overlooks the needs of older users, leading to age-related bias and a detrimental user experience (Chu et al., 2022). On the other hand, trust and security apprehensions significantly contribute to the hesitance of older individuals to interact with AI and chatbots (Iancu & Iancu, 2023). Studies indicate that a considerable segment of this demographic not only lacks the skills required to use computers proficiently but are also reluctant to provide personal health information online due to data privacy concerns and potential security breaches (Carrasco-Dajer et al., 2024; Iancu & Iancu, 2023; Shandilya & Fan, 2022; Sriwisathiyakun & Dhamanitayakul, 2022). Addressing these concerns is essential to creating an inclusive digital health environment that enhances older adults’ engagement with AI and chatbots, improving health outcomes. As healthcare providers increasingly adopt digital technologies, elderly patients need customized solutions that enhance their digital engagement and access to essential services, particularly in a post-COVID environment.

Theoretical foundation

The progression of technology has transformed various facets of everyday life, with significant impacts on healthcare and the dissemination of information. As individuals increasingly utilize digital applications, understanding the motivations and challenges affecting technology adoption is essential. Several theoretical frameworks have emerged to elucidate this phenomenon, clarifying the intricate link between people and technology. Examining how older adults interact with digital health technologies, including AI and chatbots, represents a crucial research domain, especially as such technologies become increasingly integral to healthcare. To gain deeper insights into older adults’ utilization of AI and the factors shaping their adoption of AI-driven healthcare chatbots, this study introduces a comprehensive model grounded in the expectancy confirmation model (ECM) and the unified theory of acceptance and use of technology (UTAUT2), providing a framework to analyze their engagement with these innovations.

Conceptual model and hypotheses development

The proposed research model identifies key antecedents of chatbot utilization, primarily sourced from the ECM and UTAUT2 frameworks, with perceived humanness (HPU) serving as a mediating factor. This results in a holistic model that extends the analytical robustness and applicability of the UTAUT2 framework. It integrates several UTAUT2 variables, including social influence, perceived hedonic motivation, performance expectancy, facilitating conditions, and effort expectancy, to provide a comprehensive understanding of chatbot adoption. HPU is hypothesized to mediate the relationship between satisfaction with AI-driven healthcare chatbots and the continuous intention to use chatbots among elderly individuals in Saudi Arabia. The integration of UTAUT2 with the ECM provides a synergistic framework that improves our comprehension of technology acceptance and continuation behavior, especially regarding the utilization of chatbots by older persons. It also recognizes older users’ varied and frequently intricate motivations, whose interaction with technology may be shaped by both practical needs (as delineated by UTAUT2) and emotional considerations (as outlined by ECM). By integrating variables from ECM—such as user satisfaction and confirmation—into the UTAUT2 paradigm, we can clarify how many factors interact to influence the continuous usage intentions of older individuals with chatbots. Figure 2 shows the research model.

Perceived hedonic motivation

Perceived hedonic motivation, often described as the enjoyment derived from technology use, plays a critical role in shaping its acceptance and utilization. It captures an individual’s feelings, such as enjoyment or happiness, that arise from using the technology (Venkatesh, Thong & Xu, 2012). Prior research has shown that when consumers see technology as enjoyable, their overall satisfaction and likelihood of continuing use significantly rise (Al-Azawei & Alowayr, 2020; Meske, Junglas & Stieglitz, 2019). Holbrook & Hirschman (1982) argued that hedonic motivations are essential in terms of shopping motivation and the benefits obtained from shopping. Furthermore, Meske, Junglas & Stieglitz (2019) conducted a study to investigate the influence of hedonic motives on the adoption of enterprise social networks, demonstrating that such motivations can foster sustained user engagement and satisfaction. Highlighting the importance of hedonic motivation and enjoyment in using chatbots, De Cicco et al. (2021) hypothesized that it may enhance overall user satisfaction. Building on the evaluated data, this study proposes the following research hypothesis:

• H1: Perceived hedonic motivation positively influences older adults’ satisfaction with AI-powered healthcare chatbot applications.

Facilitating condition

Facilitating conditions refer to users’ perception of the adequacy of technical infrastructure required to effectively support the successful use of a particular system. Liébana-Cabanillas et al. (2018) assert that this term embodies the belief that necessary technological and organizational infrastructures exist to enable the utilization of a specific technology (Venkatesh, Thong & Xu, 2012). Research conducted by Amer, Almahri & Bell (2020) indicated that favorable facilitating settings positively influence university students’ intentions to utilize chatbots in the future. This highlights the significance of infrastructure preparedness and support in facilitating technology adoption. By establishing the necessary technological and organizational components, institutions can enhance user confidence and augment the probability of technology acceptance across various demographic groups. This study explicitly investigates older users’ perceptions of the sufficiency of the infrastructure facilitating the usage of healthcare chatbots. Understanding whether this demographic feels the necessary resources and frameworks are available is crucial for forecasting their satisfaction with these technologies. The following research hypothesis can be proposed:

• H2: Facilitating conditions positively influences older adults’ satisfaction with AI-powered healthcare chatbot applications.

Social influence

Social factors play an important role in shaping user’s evaluations of technology’s usefulness (Athota et al., 2020; Tubaishat, 2018). Previous research elucidates how people change their beliefs and behavior to fit into those of people around them in a bid to gain acceptance (Tan & Liew, 2022). Consistent with this assertion, Shen et al. (2006) demonstrated that social influence or perceived social pressure significantly impacts technology acceptance, meaning that when individuals observe their role models embracing a particular technology, they are likely to view this as a worthy venture due to its usefulness. Khadija, Zahra & Naceur (2021) observed that peer recommendations affect the perception of technology in the healthcare context, specifically, if patients see others using and endorsing chatbots, they are likely to regard these chatbots as beneficial for health management. In the realm of mobile learning, Criollo-C et al. (2021) assert that social influence exerts a diminished effect on the acceptance of mobile learning. On the contrary, Bonn et al. (2016) emphasized that social factors significantly influence individuals’ perceptions and behavioral intentions about technology usage. Perceived social pressure or support can lead to heightened satisfaction as users feel more integrated into their social networks (Pal, Funilkul & Vanijja, 2020). For older adults, social influence and support from family, caregivers, or community members can substantially augment their confidence and satisfaction with AI-powered healthcare chatbots. The following research hypothesis can be proposed:

• H3: Social influence positively influences older adults’ satisfaction with AI-powered healthcare chatbot applications.

Performance expectancy

Performance expectancy refers to an individual’s belief that using a particular technology will enhance their performance or effectively meet their needs, as defined by Venkatesh et al. (2003). This construct plays a pivotal role in shaping both the acceptance of a technology and the satisfaction users derive from its use. According to Choudhury & Shamszare (2024), a favorable assessment of a chatbot’s efficacy in executing its designated functions can result in heightened user satisfaction. Furthermore, Abd Aziz, Aziz & Abd Rahman (2023) underscore the notion that satisfaction can function as a significant intermediary variable that mediates the effects of performance expectancy on the comprehensive technological experience. Rahi et al. (2019) similarly stressed the critical nature of performance expectancy. The authors posit that when users perceive elevated performance levels, this perception correlates with enhanced satisfaction outcomes. In the context of healthcare applications, Sitthipon et al. (2022) discovered that users exhibit a greater propensity to engage with healthcare chatbots and applications when they believe that these technological tools will facilitate improved access to healthcare services and optimize overall health management. Based on the evaluated data, this study proposes the following research hypothesis:

• H4: Performance expectancy positively influences older adults’ satisfaction with AI-powered healthcare chatbot applications.

Effort expectancy

Effort expectancy refers to the degree of simplicity and user-friendliness associated with interacting with a specific technology (Venkatesh, Thong & Xu, 2012). It encompasses the adaptability and user-friendliness of technology from an individual perspective. Empirical studies indicate that effort expectancy significantly affects users’ intentions to embrace novel technologies. For example, Polyportis & Pahos (2024) underscore that effort expectancy is associated with a strong positive impact on students’ ChatGPT use behavior. Sari et al. (2024) also concluded that not only performance but also effort expectancy are crucial determinants in the technology adoption process among Moroccan nursing students. Furthermore, Elkhatibi, Guelzim & Benabdelouahed (2024) researched the variables impacting the adoption of AI chatbots in the banking sector, affirming that elevated levels of effort expectancy are positively correlated with user satisfaction. The investigation carried out by Elok & Hidayati (2021) provides valuable insight into the close interconnection between effort expectancy and user satisfaction regarding digital wallet utilization. Their findings emphasize that effort expectancy constitutes a vital variable influencing user satisfaction. This association is particularly noteworthy within the paradigm of user engagement with digital platforms, such as AI-enhanced healthcare chatbots. Building on the evaluated data, this study proposes the following research hypothesis:

• H5: Effort expectancy positively influences older adults’ satisfaction with AI-powered healthcare chatbot applications.

Confirmation

The influence of confirmation on the satisfaction derived from new technology constitutes a pivotal domain of inquiry, particularly as users increasingly interact with sophisticated solutions such as artificial intelligence and digital platforms. As articulated by Bhattacherjee (2001), the affirmation of expectations has a substantial impact on sustained satisfaction and the continued utilization of technological resources. Within the realm of digital commerce, Guo, Zhang & Xia (2023) examined the implications of website design on customer satisfaction and loyalty. They contend that when users perceive that their experiences align with or surpass their prior expectations, there is a corresponding increase in satisfaction, which fosters enhanced loyalty. Furthermore, Sinha & Singh (2023) conducted a focused investigation into the engagement of elderly consumers with multichannel banking services, reevaluating the ECM to demonstrate the significance of confirmation in determining satisfaction levels. Their research suggests that for older adults, who may encounter unique usability challenges, the affirmation of expectations is a critical determinant of their overall satisfaction. Additionally, Thong, Hong & Tam (2006) emphasized that post-adoption beliefs, which stem from the confirmation of initial expectations, are essential for maintaining user engagement with information technology. This relationship is further corroborated by the research of Cavallo et al. (2018), who investigated robotic services for older adults, discovering that perceived usefulness and satisfaction are intricately linked to the extent to which users’ expectations are validated through their experiences. Analyzing mobile instant messaging applications, Oghuma et al. (2016) found that the confirmation of expectations significantly influences users’ intentions to persist in utilizing these services. Consequently, we postulate the following hypothesis:

• H6: Confirmation positively influences older adults’ satisfaction with AI-powered healthcare chatbot applications.

Perceived security

Perceived security refers to the degree to which users believe their interactions with technology such as chatbots are secure and devoid of any inherent risk predispositions (Shrestha et al., 2021). Notably, the use of chatbots poses significant security threats since they frequently handle sensitive data, necessitating the implementation of rigorous security measures (Keerthana, Vaishali & Anitha, 2024). Nicolescu & Tudorache (2022) argue that threats associated with personal data protection are regarded as one of the key challenges and constraints related to the utilization of chatbots for both organizations and users. The security of healthcare chatbots is critical to protecting sensitive data, hence fostering trust among users (May & Denecke, 2022). When the user believes that his or her information stored on the personal and medical data is protected, the user will not have any reservations about using the chatbot as it is deemed a safe mode of communication. Additionally, implementing robust security features can eliminate users’ unauthorized access to the system and limit the occurrence of interruption overall (Athota et al., 2020). Given the potential adverse effects of privacy and security risks, particularly in critical sectors like banking and healthcare, safeguarding user data is pivotal for achieving customer satisfaction within the emerging disruptive technology landscape (Cheng & Jiang, 2020; Cox & Dale, 2001; Taehyee, Taekyung & Namho, 2020). When users are assured that their interactions with the chatbot are secure, then they will spend more of their time engaging with the application and experimenting with its functionalities, hence deepening the ease-of-use construct (Kasilingam, 2020). Building on the evaluated data, this study proposes the following research hypotheses:

• H7: Perceived security positively influences older adults’ satisfaction with AI-powered healthcare chatbot applications.

• H8: Perceived security positively influences older adults’ continuous intention to use AI-powered chatbots.

Satisfaction and continuous intention to use AI-powered chatbots

Satisfaction with technology markedly affects users’ intentions to persist in utilizing a given product or service. Empirical research substantiates the assertion that favorable user experiences arising from satisfaction are directly associated with prolonged engagement. For example, Alruwaili, Shaban & Elsayed Ramadan (2023) conducted a systematic review centering on digital health interventions and their contributions to the promotion of healthy aging. As the authors highlight that user satisfaction plays a critical role in ensuring the continued use of digital health tools, they note that when individuals recognize these technologies as beneficial and effective, they are more likely to utilize them consistently. Cheng & Jiang (2020) also explored the implications of AI-driven chatbots on user experience. Their results indicate that satisfaction, influenced by the perceived gratifications derived from chatbot interactions and the concomitant privacy concerns, significantly impacts users’ loyalty and intent to continue employing these technologies. When users report satisfaction with their interactions, they exhibit a greater propensity to remain loyal to technology and to utilize it repeatedly. The following research hypothesis can be proposed:

• H9: Satisfaction with AI-powered healthcare chatbot applications positively influences older adults’ continuous intention to use AI-powered chatbots.

The mediation effect of perceived humanness

The Humanizing Experience Theory underscores the significance of fostering a perception of humanity within digital interactions, especially via the use of chatbots (Singh & Malik, 2024). Perceived humanness is a construct commonly used in human–computer interaction (HCI), defined as the extent to which an entity is regarded as human (Hendriks et al., 2020). Multiple studies have reinforced the importance of human-like qualities in accepting chatbots (Rapp, Curti & Boldi, 2021; Hendriks et al., 2020; Go & Sundar, 2019; Elson, Derrick & Ligon, 2018). Go & Sundar (2019) contended that chatbot attributes, like friendly interface and natural language, can improve users’ impressions of the chatbot as more anthropomorphic. This humanization can cultivate emotional connections, resulting in enhanced satisfaction and sustained utilization. Lei, Shen & Ye (2021) conducted analogous research that focused on the perceived significance of humanness in chatbot interactions. The authors revealed that users frequently associate a more anthropomorphic appearance with elevated satisfaction levels. This indicates the imperative for chatbot designers to integrate human-like characteristics to fulfill user expectations and improve the overall experience. Further highlighting the importance of chatbot humanization in customer service, Hsu & Lin (2023) posited that consumers’ satisfaction and loyalty towards customer service chatbots are associated with perceived humanness. On the other hand, Ding & Najaf (2024) emphasized that chatbots with human-like characteristics might foster more engaging and relatable interactions, hence improving user satisfaction and increasing the probability of continuing usage. Chattaraman, Kwon & Gilbert (2012) investigated the impact of human-like interaction in virtual agents on retail websites, with a particular emphasis on elderly consumers. Their research demonstrated the significance of building virtual agents with human-like attributes to serve elder users and improve their satisfaction efficiently. As older users exhibit diverse degrees of comfort and familiarity with technology, rendering the human-like attributes of chatbots is crucial for enhancing happiness and promoting sustained usage (Chattaraman, Kwon & Gilbert, 2012). Through the integration of anthropomorphic attributes, including empathy and distinct personality traits, chatbots can potentially enhance user interaction engagement. Building on the evaluated data, this study proposes the following research hypothesis:

• H10 : Perceived humanness serves as a significant mediator in the relationship between user satisfaction and the continuous intention to utilize these technologies among elderly users.

Table 1 provides summary for research hypotheses.

Research design

Healthcare chatbot systems offer personalized and effective health recommendations, with accuracy that closely resembles human medical expertise. Accordingly, this research seeks to identify the elements that encourage older patients to consistently engage with AI-driven healthcare chatbot technologies. The conceptual framework integrates perspectives from motivation theory, UTAUT, and ECT to provide a comprehensive understanding of these factors. To achieve the research objectives, a quantitative approach was adopted, focusing on the continuous usage of AI-powered healthcare chatbots. Data were gathered through a structured survey questionnaire, a method well-suited for systematically capturing behaviors, attitudes, perceptions, and beliefs, thereby enabling the extrapolation of findings.

Data collection procedures

An exploratory, cross-sectional survey design was employed to examine the adoption of AI-driven healthcare chatbot solutions among elderly individuals in Saudi Arabia. This research was carried out under the auspices of the Faculty of Economics and Administration at King Abdulaziz University. The Research Ethics Committee (REC) reviewed and approved the research (reference number: REC 1/5). Participants provided informed consent before participation to confirm their voluntary involvement and awareness of data confidentiality. Data collection was conducted through an anonymous survey distributed via Google Forms; an online platform known for its efficiency in managing electronic surveys. Moreover, Google Forms has a user-friendly design and robust data management capabilities. In alignment with our framework, participant eligibility was restricted to those aged 60 and above who utilized healthcare chatbots solutions. A convenience sample was obtained using a snowball sampling technique facilitated through WhatsApp, the most commonly used social networking tool in Saudi Arabia, during June and July 2024. The researchers utilized their professional and personal networks to disseminate the survey invitation. All responses were anonymized, and no personally identifiable data were collected.

This sampling method was chosen for two primary reasons. First, WhatsApp’s widespread use in Saudi Arabia makes it an ideal platform for reaching the target demographic. Second, the snowball approach enabled the study to reach older participants by leveraging familial and social connections. Interested respondents were guided to the survey platform, where they could review the study’s objectives.

Correspondingly, respondents were required to provide their consent and confirm their age prior to enrolling in the survey. The survey invitation explicitly stated that the participation is voluntary and participation in the survey will imply providing informed consent to guarantee the confidentiality of participants. Participants indicated their consent by selecting “Agree” to proceed to the survey questions.

A total of 162 responses were collected, with a response rate of 74%. After removing four submissions due to missing information or uniform responses across all questionnaire items, the final sample size comprised 158 participants, representing 71% of the initial pool. Participant anonymity was assured, and no answers were deemed right or wrong. The survey design prioritized clarity and precision, with ambiguous terminology eliminated to ensure comprehensibility and accuracy in responses.

Data analysis techniques

Partial least squares structural equation modeling (PLS-SEM) was employed for data analysis, employing the Smart PLS 4.0 software suite (Silva, Shojaei & Barbosa, 2023; Cheah et al., 2018). This method, which applies partial least squares estimation, is particularly suitable for handling complex models comprising multiple constructs and indicators (Guenther et al., 2023). A key advantage of PLS-SEM is its ability to analyze data effectively without requiring a large sample size due to its component-based approach (Lohmöller, 2013).

Survey instruments

To ensure methodological rigor and relevance to the study’s context, the measurement scales for each construct were adapted from established scholarly sources. For instance, ‘Perceived Security’ and ‘Perceived Hedonic Motivation’ were assessed using three items each, derived from Noor, Rao Hill & Troshani (2022) and a combination of Iancu & Iancu (2023) and O’Brien (2010), respectively. Four items measuring ‘Facilitating Condition’ were informed by the works of Venkatesh et al. (2003), Venkatesh, Thong & Xu (2012). ‘Social Influence’ relied on two items adapted from Venkatesh et al. (2003), Venkatesh, Thong & Xu (2012), while ‘Performance Expectancy’ was assessed with three items sourced from Ni et al. (2024) and Venkatesh et al. (2003), Venkatesh, Thong & Xu (2012). ‘Effort Expectancy’ included two items based on Slade, Williams & Dwivedi (2013) and Venkatesh, Thong & Xu (2012). Similarly, ‘Confirmation’ utilized three items from Aggarwal et al. (2023) and Lee, Hsieh & Hsu (2011), and ‘Satisfaction’ was measured with four items following Bhattacherjee (2001). ‘Perceived Humanness’ included three items adapted from Go & Sundar (2019), while ‘Continuous Intention & Immersion to use AI-Powered Chatbots’ employed three items primarily based on Al-Sharafi et al. (2023) and Thong, Hong & Tam (2006). All constructs were measured using a 5-point Likert scale ranging from 1 (strongly disagree) to 5 (strongly agree). Further details on the measurement instruments and their sources can be found in Appendix A.

Demographic information

The demographic data presented in Table 2 reveals that 46.2% (n = 73) of respondents were male, while 53.8% (n = 85) were female. Notably, 67.7% (n = 107) of participants reported having bachelor’s degrees, indicating a predominantly highly educated cohort. The statistical analysis of income distribution indicates that a notable proportion of participants, specifically 46.8%, receive annual earnings that are less than 16,000 SAR. Simultaneously, 29% of the individuals surveyed are categorized within the moderate-income bracket, which ranges from 16,000 to 26,000 SAR, whereas 15.8% of respondents indicate that their earnings surpass 26,000 SAR. Furthermore, 2.5% of the respondents opt not to disclose their income information, and 5.7% express uncertainty regarding their income levels. Regarding their reasons for using healthcare chatbots, 31% (n = 49) utilized the technology for medical assistance, 30.4% (n = 48) for prescription refills, and 5.1% (n = 8) for scheduling appointments. Finally, 54.4% (n = 86) of the participants indicated that they accessed healthcare chatbot from their smartphones, and 45.6 % (n = 72) used their personal computers to access healthcare chatbot.

Measurement model assessment

This research employed PLS-SEM as a data analysis tool to evaluate convergent validity, discriminant validity, and reliability (Hair et al., 2019). PLS-SEM, a variance-based approach, is particularly suited for estimating structural models and optimizing explained variance in dependent variables (Aburumman et al., 2022). The evaluation included internal consistency reliability through Cronbach’s alpha (CA) and composite reliability (CR). As shown in Table 3, all CA values exceeded 0.60, and CR values surpassed 0.70, indicating robust internal consistency (Hair et al., 2019). A Cronbach’s alpha value of 0.60 or higher was regarded as acceptable for each construct’s internal consistency. Convergent validity was assessed via average variance extracted (AVE), with all constructs achieving AVE values greater than 0.50, confirming the scales’ validity and reliability.

Discriminant validity

To ensure constructs were distinct, the study assessed discriminant validity using the Fornell–Larcker criterion and Heterotrait-Monotrait Ratio (HTMT). Table 4 shows that HTMT values were below 0.85, confirming adequate differentiation between constructs (Aburumman et al., 2022). Furthermore, Table 5 demonstrates that the square root of each construct’s AVE exceeded its correlations with other constructs, affirming discriminant validity.

Model fitness

Model fitness, reflecting the alignment of the conceptual model with observable data, was assessed using SRMR. The SRMR value of 0.077, below the threshold of 0.08, confirms satisfactory model fitness (Hair et al., 2021). As detailed in Table 6, R² values indicate that independent variables explained 76.5% of variance in satisfaction with AI-powered chatbots, while satisfaction, perceived security, and perceived humanness accounted for 62.2% of the variance in continuous intention to use chatbots. Perceived humanness alone contributed to 39.7% of variance in satisfaction. These findings validate the model’s predictive power and accuracy (Miles, 2005).

Structural model assessment

This study employed the bootstrapping technique, generating 5,000 sub-samples to evaluate the research hypotheses and assess the statistical significance of path coefficients (Hair et al., 2019). The structural model was then analyzed to test the hypotheses and calculate effect sizes (Hair, Howard & Nitzl, 2020). This assessment focused on determining the model’s predictive accuracy and the relationships among constructs. A 5% significance level with a 95% confidence interval was applied, requiring p-values below 0.05 and t-values above +1.96. Table 7 provides the results of hypothesis testing, and Fig. 3 presents the structural equation model.

H1 confirms a significant positive relationship between perceived hedonic motivation and satisfaction with AI-powered chatbots (β = 0.110, t = 1.972, p = 0.049). Similarly, H2 indicates that facilitating conditions significantly enhance satisfaction (β = 0.190, t = 2.077, p = 0.038). In contrast, H3 shows that social influence does not significantly impact satisfaction (β = 0.089, t = 1.742, p = 0.081). Performance expectancy exerts a strong positive effect on satisfaction (H4: β = 0.262, t = 3.570, p = 0.000). Effort expectancy and confirmation also demonstrate positive influences (H5: β = 0.164, t = 2.625, p = 0.009; H6: β = 0.158, t = 1.996, p = 0.045). Perceived security significantly impacts both satisfaction (H7: β = 0.287, t = 2.985, p = 0.003) and continuous intention to use AI-powered chatbots (H8: β = 0.414, t = 4.612, p = 0.000). Satisfaction positively influences continuous intention (H9: β = 0.207, t = 2.003, p = 0.045). Finally, H10 reveals that perceived humanness mediates the relationship between satisfaction and continuous intention (β = 0.184, t = 3.117, p = 0.002). Collinearity diagnostics confirmed that all variance inflation factor (VIF) values were below 3.3, indicating no multicollinearity issues (Hair et al., 2021; Miles, 2005).

Theoretical implication

The present work significantly contributes to scientific literature in the healthcare and technology context. This research offers three notable contributions to the theoretical understanding and modeling of technology acceptance. First, the findings provide valuable enhancements to the theoretical foundation concerning new technology adoption. By leveraging insights from motivational theory, UTAUT, and ECT, this study examines the key factors influencing the sustained use of intelligent healthcare chatbots by older adults. Second, the study identified several factors including perceived security, perceived humanness, performance expectancy, and effort expectancy that significantly enhanced existing literature, indicating that more human-like interactions can foster a stronger emotional connection, resulting in increased satisfaction and continued use of AI-powered healthcare chatbots among older adults. Finally, the results underscore the importance of regional context, particularly in Saudi Arabia. The study offers empirical evidence from both demographic and geographic perspectives that have hitherto been neglected. By situating the findings within Saudi Arabia’s unique cultural and socio-economic context, this study lays the groundwork for future investigations into the adoption of emerging technologies in similar settings. This regional emphasis strengthens the theoretical foundation of technology acceptance. It offers valuable avenues for future research that can inform practical applications in healthcare technology and act as a comparative benchmark for further research in other emerging markets.

Practical implications

The study results inform actionable recommendations for organizations and policymakers aiming to utilize AI-powered healthcare chatbots to improve older adults’ engagement and satisfaction. First, key predictors of satisfaction and continued use, identified in the study, underscore the importance of designing chatbot interfaces that prioritize user accessibility, interactivity, and clarity. Healthcare providers should focus on creating chatbots that seamlessly integrate into older adults’ existing digital habits and preferences, promoting both ease of use and enjoyable experience. Second, the study highlights that social influence does not significantly affect the intention to continue using AI-powered chatbots. This suggests that marketing strategies should shift away from emphasizing social aspects and instead focus on individual benefits and direct user experiences. Highlighting the functional benefits of chatbots, particularly their capacity to deliver precise and timely health information, can prove more impactful in encouraging user adoption. Third, healthcare organizations should prioritize the development of AI-powered chatbot with intuitive, user-friendly, interfaces. Simplifying navigation and minimizing the steps required to access information can significantly enhance user experience, making the technology more approachable for older adults. Similarly, ensuring that chatbots are easily integrated into existing healthcare systems and workflows can reduce the perceived complexity for both healthcare providers and patients. Seamless integration with electronic health records and other digital health tools can facilitate easier access to information and improve overall sustainability.

Research limitations and future directions

Despite its significant theoretical and managerial contributions, this study acknowledges several limitations. Namely, the study’s sample size of 158 healthcare chatbot users aged over 60 from Saudi Arabia, while informative, might not reflect the full range of diversity within the broader elderly population. Future investigations should consider utilizing larger sample sizes to improve the applicability of the findings across other countries in the region. Additionally, the reliance on a convenience sample distributed via WhatsApp introduces potential selection bias, as older adults who are more digitally literate or comfortable with mobile technology may be overrepresented. Since this sampling approach may exclude individuals with lower digital proficiency, it limits the generalizability of the findings to all elderly populations. From another standpoint, the dependence on self-reported data could introduce biases, such as recall inaccuracies or the inclination to provide socially desirable responses, which may affect data reliability. With these biases being potentially further compounded by the self-selection effect inherent in digital survey distribution, careful interpretation of results is warranted. To address these limitations, future studies should explore alternative recruitment methods by including offline surveys and hybrid data collection strategies.

Furthermore, cross-sectional nature of this research restricts the ability to draw conclusions about causative relationships between the studied factors and the ongoing use of healthcare chatbots. To address this, longitudinal research designs are recommended for future studies to better capture the dynamic relationships over time. Experimental approaches could further enhance understanding by isolating causal effects of chatbot adoption through controlled interventions. One of the primary recommendations is to measure behavioral changes in response to specific chatbot features or personalized AI-driven interactions.

As such, future scholars must consider incorporating mixed-methods approaches, combining qualitative and quantitative data, to address these biases and offer a more well-rounded perspective on user behavior. The study’s scope is also limited to certain variables, potentially overlooking other significant factors that could influence chatbot use. Future research could examine additional influencing factors, including but not limited to privacy concerns, trust, and health literacy. Studies exploring AI-driven chatbot interventions in clinical settings or comparing their effectiveness across different healthcare models could provide deeper insights into their real-world applicability and patient outcomes. Investigating the prolonged effects of chatbot usage and the practical difficulties associated with integrating AI technologies into healthcare systems could also provide deeper insight into the associated benefits and challenges.