The information ecology theory was proposed by Horton [10] in the 1980s, emphasizing the harmonious symbiotic relationship among “information human, information, environment, and technology” in the information system, and focusing on the interdependence, interaction, and dynamic evolution process among various elements of the system. Among them, the information person is the main body of the system, leading the initiation and interaction of information activities. Information exists as an object, carrying knowledge, data, and meaning. Technology provides tool support for the processing and transmission of information. The environment constitutes the field where information human, information, and technology interact, shaping the patterns and boundaries of information behavior [11]. Subsequently, Davenport [12], and Nardi and O’Day [13], among others, further developed this theory, defining it as “the relationship between people and organizations in the information environment,” viewing “the information ecosystem” as “an organic system of people, practices, values and technologies in a specific environment,” and pointing out that the information ecology refers to a dynamic equilibrium state achieved through the transmission and feedback activities of information human and the information environment within a specific information space, using information technology as a means [14]. This theory emphasizes the application of ecological methods to study the interaction mechanisms among system factors [15], with the goal of achieving “information ecological balance” [16]. In conclusion, the information ecosystem is an open and organic whole with human beings as the core element, and there exists an information flow relationship among these elements.

The existing technology acceptance model focuses on linear cognitive drive, while the SERVQUAL (service quality theory) concentrates on the gap between expectations and perceptions. Both of these approaches are unable to explain the nonlinear interaction and multifactor synergy in human-AI collaborative medical consultation. The social technology system theory, although focusing on the interaction between social and technological subsystems, lacks a sufficient depiction of “information” as an independent element and external environmental norms. The institutional perspective mainly emphasizes the macrolevel and is difficult to integrate with the microperception variables of individual patients. The information ecology theory, on the other hand, provides a more macro and dynamic systemic lens, emphasizing the interdependence, coevolution, and adaptive configuration among the four elements of “information human, information, technology, and environment.” The human-AI collaborative medical consultation process in OHCs is essentially a process of information flow and reconfiguration, constituting a typical information ecosystem [17]. In this system, patients act as information human to initiate interactions, diagnostic information flows as the core object, intelligent systems and platforms provide technical support, and regulations and ethical norms form the environmental field. All four elements are present. The consultation process presents a continuous information cycle: patients describe symptoms, AI assists in triage, doctors provide diagnosis feedback, patients generate efficacy feedback after execution, and this triggers new information needs, forming a multidirectional interaction network. These elements are deeply coupled and interact with each other, jointly determining the effectiveness of the consultation service.

Based on this theoretical perspective, the core variable system constructed in this study is not a simple superposition of existing theories, but a systematic logical reconstruction: the classic constructs of technology acceptance and service quality are respectively anchored in the “information technology” and “information person” dimensions, and the attribute of information is independently classified as the “information” dimension and incorporated into the ecological regulatory field of “environment.” This design transcends the explanatory boundaries of traditional linear models and, by examining the collaborative configuration of “person-information-technology-environment,” reveals the complex causal chain behind the effectiveness of human-AI collaborative medical consultation, providing a more systematic theoretical basis for optimizing the digital health service ecosystem.

As an important tool for integrating multiple resources and enhancing the effectiveness of medical services, OHCs have transcended traditional information exchange spaces and become platforms that support users in sharing experiences, consulting experts, and promoting knowledge cocreation [18-20]. With the help of Web 2.0 technology, these communities provide services such as health and medical information, online consultation, and knowledge question and answer, building a virtual communication space for patients, doctors, and health enthusiasts [21,22]. Human-AI collaboration aims to establish a deeply integrated human-AI interaction model by integrating the advanced cognitive abilities of humans with the technical advantages of machines, achieving complementary strengths and efficient collaboration [23,24]. This system framework emphasizes handing over the final decision-making power to personnel with professional knowledge and experience, in order to overcome model uncertainties and ensure the reliability and stability of decisions [25]. In medical human-AI collaboration, AI systems actively identify user preferences, analyze vast medical literature and clinical cases, and accordingly provide diagnosis and treatment suggestions to doctors. Doctors then exercise their subjective initiative to evaluate, modify, or appropriately adopt these suggestions [26,27]. Through this dynamic coordination and efficient cooperation, machines focus on handling structured tasks, while humans concentrate on higher-level judgments and innovations. Both parties exchange information through specific interaction methods to jointly achieve work goals, thereby effectively improving service quality and work efficiency, and achieving the core goal of reducing burden and increasing efficiency [28-30]. In conclusion, the existing literature mostly focuses on the technical aspects of human-AI collaboration mechanisms, but pays insufficient attention to the specific context of OHCs. Therefore, this paper defines the human-AI collaborative consultation in OHCs as a process in which patients interact with AI systems and professional doctors on OHC platforms to obtain medical services such as health consultation and disease diagnosis. In this mode, patients input information through the platform. The AI system uses big data and algorithms to initially analyze and screen the symptoms, medical history, and other information input by the patients, and then provides auxiliary diagnostic suggestions for doctors. Doctors make a comprehensive judgment based on their professional knowledge and provide the final diagnosis and treatment plan.

Scholars have comprehensively used various qualitative and quantitative methods to systematically explore the influencing factors of PS in the human-AI collaborative consultation model, revealing its multidimensional and interactive complexity. In terms of quantitative research, relevant results rely on multiple theoretical models to clarify the interrelationships of key variables. The structural equation model reveals that functional value, social value in perceived value, as well as interactivity, perceived quality, etc, have a positive effect on satisfaction [4,31], while information accuracy (IA) and privacy security can indirectly affect satisfaction by influencing perceived benefits and costs [32]; Zhai et al [33] further subdivided the quality dimensions and verified that personnel quality, procedural quality, technical quality, and cost perception have differentiated effects on the satisfaction of remote consultation patients. Based on the unified theory of acceptance and use of technology model, the research focused on highlighting the core values of performance expectations, effort expectations, and social influence [34,35]. Meanwhile, related studies such as the perception risk model, the service quality model, and the trust theory have also confirmed the crucial restrictive and driving roles of perceived risk and doctor-patient trust in the consultation service [6,8]. Furthermore, studies from an information ecosystem perspective have found that environmental elements such as community interaction, atmosphere, and platform governance norms can effectively boost patients’ willingness to pay for consultation [22,36,37]. In OHC consultations, doctor-patient interaction and technical factors are particularly crucial. Wu et al [5] based on the dual-factor theory confirmed that the information support and emotional support provided by doctors are indispensable, and Deng and Hong [8] further emphasized the impact of website credibility. In qualitative and text analysis methods, different research methods have also formed complementary conclusions. Grounded theory related research has extracted multiple factors such as environment, platform, and physician services, which jointly affect user evaluation [38,39]; latent Dirichlet allocation model and Kano model further confirm the core roles of information quality, system quality, service quality, and privacy protection [9,21,40-42]; text mining analysis takes a communication perspective and confirms that the amount and comprehensibility of doctors’ responses significantly enhance satisfaction [43]. Meta-analysis–related results also verify the stable influence of the structural quality, process quality, and outcome quality of remote medical services [44].

In summary, although the existing studies have approached from multiple theoretical perspectives and widely identified various influencing factors, their common limitation lies in focusing on the independent net effect of a single or a few factors. This paradigm often presupposes linear relationships and independence between variables, resulting in a lack of a systematic integration framework that considers the four elements of information, information human, technology, and environment as an organic whole, and failing to deeply explore how multiple factors interact and collaborate dynamically in the specific scenario of “human-AI collaboration” to jointly affect consultation efficacy. It is difficult to reveal the possible multiple concurrent configuration paths for achieving high satisfaction, limiting the precision and situational adaptability of optimization strategies.

In view of this, this study introduces the perspective of information ecology and constructs an integrated analysis framework covering four dimensions: information human, technology, information, and environment. Focus on the analysis of the interactive relationship and collaborative mechanism of the four core elements during the consultation process; by applying fsQCA to explore the multiple concurrent condition combinations and equivalence realization paths that affect the effectiveness of human-AI collaborative consultation, we aim to break through the limitations of single-factor linear analysis and provide systematic theoretical support and practical guidance for precisely optimizing the service effectiveness of OHCs and promoting their sustainable development.

This paper first conducts a literature analysis to sort out the relevant research, identifying various potential influencing factors under the four core dimensions of information human, information, technology, and environment. After expert consultation for revision and verification through practical cases, 13 specific influencing factors were ultimately determined, as shown in Table 1. Based on the essence of the information ecology theory and the scenario characteristics of human-AI collaborative consultation, it can be seen that the 4D elements do not exist independently but interact through complex relationships to jointly affect PS and the efficiency of the consultation service. Therefore, this paper will focus on the above four core dimensions to deeply analyze the connotations and mechanisms of action of each specific influencing factor, and construct a systematic theoretical model to reveal the formation logic and influence path of PS in the OHC’s human-AI collaborative consultation model.

The information human factor reflects the subjective cognition and emotional response of patients in human-AI collaborative consultation, directly influencing their behavioral decision-making and consultation effectiveness. This study clearly defines the “information human” as the core entity that engages in information interaction, generates information needs, and perceives the consultation services in the human-AI collaborative consultation within OHCs. This entity is the patient who participates in the consultation and is also a user of this type of health service. To maintain the rigor and readability of the text, the following terminology guidelines will be followed: when presenting the theoretical model, “information human” will be used; while “user” or “patient” will only be used when referring to the research participants or the actors, to avoid confusion with the aforementioned core constructs. Patients’ concerns about potential risks such as medical quality, privacy protection, misdiagnosis risk, and waste of time or financial costs constitute perceived uncertainty (PU), significantly reducing trust and willingness to use. Perceived service effectiveness (PSEF), perceived service accuracy (PSA), and perceived service empathy (PSE) represent patients’ value judgments on core dimensions of system service quality, such as response speed, result reliability, and personalized care, and actively shape patients’ satisfaction and willingness to participate [31,45,46].

PU arises from concerns regarding the incompleteness of regulations and the immaturity of technologies associated with online consultation. It manifests as the perception of risks such as misdiagnosis, privacy leakage, and waste of time or money [35,47]. PSEF refers to a patient’s perception of the system’s ability to provide the required services quickly and effectively. PSA is the patient’s subjective judgment of the degree of alignment between the consultation service’s output and the patient’s own health needs. PSE refers to a system that delivers personalized and targeted services and care based on the patient’s medical condition and personal circumstances, thereby making the patient feel understood and valued. Collectively, these perceptions constitute the key psychological mechanisms underlying patients’ behavioral decisions, profoundly reflecting the subjectivity and initiative of the information human and its core driving role in human-AI collaborative interaction.

In online health consultation services, users’ adoption of inappropriate information may lead to serious consequences. Therefore, providing high-quality and highly persuasive information is of vital importance. Information quality is the key variable affecting user satisfaction [7] and encompasses three dimensions: information usefulness (IUS), IA, and information understandability (IUD). Together, these three dimensions form the basis for users to evaluate the value of information, and by influencing patients’ satisfaction, trust, and behavioral decisions, they shape the effectiveness of information adoption, usage, and dissemination, thereby sustaining the service quality and the overall viability of the information ecosystem in online human-AI collaborative consultation.

IUS refers to the perceived utility of the information or diagnostic results obtained by patients during the human-AI collaborative consultation process. IUS meets the core demand goals of users, which is the direct driving force for their continuous use and satisfaction. When patients truly experience the alleviation of health problems through human-AI collaborative consultation, such as reduced symptoms and decreased anxiety, the perceived usefulness of the information will significantly enhance their satisfaction and positively influence their willingness to share and continuously use the information [9,14,32,48].

IA refers to the degree of precision and reliability of the consultation information in aspects such as disease description, diagnostic conclusion, and prescription recommendation [49]. IA is the foundation for building trust and ensuring safety and reliability. Its mechanism of action is as follows: (1) accuracy of disease description: ensuring that the system precisely captures and structurally presents the details of symptoms, laying the initial trust of being understood. (2) Accuracy of disease diagnosis: relying on the deep collaboration between AI medical knowledge and doctors’ experience, a diagnosis that conforms to clinical guidelines and takes into account individual differences is produced, which triggers patients’ professional identity and strengthens their treatment expectations. (3) Prescription accuracy: ensuring the safety and effectiveness of the treatment plan through database verification and personalized adjustment, thereby enhancing patients’ confidence in the certainty of the diagnosis and treatment behavior. (4) Feasibility of suggestions: transforming professional information into action guidelines that fit patients’ life scenarios, thereby increasing their willingness to engage in health management. IA is an essential attribute for enhancing PS [21]. Higher IA directly strengthens patients’ trust in the system, confidence in treatment, and perception of service value, thereby significantly improving satisfaction.

IUD refers to the degree to which the content of the consultation information is expressed clearly and understandably [31]. IUD reduces users’ cognitive load and promotes the effective absorption of information. In the context of human-AI collaboration, if information presentation endures problems such as an accumulation of technical jargon, unclear logic, or obscure expression, it directly undermines users’ initial evaluation of service quality and leads to a decline in satisfaction. Conversely, the use of plain language, visualization tools, and similar strategies can effectively reduce the cost of understanding, enhance information acceptance, and improve PS [31].

The healthy development of online health human-AI collaborative consultation in health contexts relies on support from a stable multilevel environment. Legal and regulatory protection (LRP), platform ethical norms (PEN), and social rules (SR) interact synergistically, respectively, constructing the fundamental environment for system operation from the perspectives of rigid institutional constraints, organizational self-discipline, and sociocultural norms. These environmental factors, by shaping patients’ sense of institutional trust, organizational security, and behavioral predictability, reduce their perceived risks and jointly regulate and maintain the compliance, sustainability, and user satisfaction of the system.

LRP refers to patients’ evaluation of the fulfillment of leadership, guarantee, supervision, and management responsibilities by government departments in the medical industry. A comprehensive legal and regulatory system for internet health care, along with supporting mechanisms such as online medical dispute resolution and insurance coverage, reflects the state’s institutional commitment and governance capacity in this emerging field. Patients’ positive perception of policy effectiveness can effectively enhance their trust in the system [50], reduce concerns about systemic risks, and positively influence their evaluation of service satisfaction [51].

The PEN refers to patients’ evaluation of the system for protecting patient information security as formulated by the OHC platform. Ethical norms cover privacy protection, data security, and maintenance of user trust. The platform needs to make patients fully aware of its commitment and ability to protect personal information security through clear and transparent policy communication—such as minimizing data collection, implementing strict authorization mechanisms, and prohibiting data abuse [52]. High trust in the platform’s ethical norms can significantly alleviate patients’ perception of privacy leakage and data abuse risks, enhance their sense of organizational security, and serve as a key organizational commitment to maintaining user trust and improving satisfaction [53,54].

SRs refer to the extent to which the words and deeds of the people around the patient, as well as the influence of media publicity, affect the patient’s use of the human-AI collaborative consultation system. The pathways through which social norms influence patients are as follows: first, interpersonal influence and imitation—positive usage experiences and recommendations from relatives, friends, or similar groups can enhance patients’ confidence in using the new technologies and their expected efficacy [55,56]; second, media shaping and public opinion—positive media reports and a supportive social atmosphere created by public opinion reduce social resistance and cognitive uncertainty regarding the adoption of new technologies. Positive signals from SRs positively influence patients’ adoption willingness and satisfaction evaluations by reducing their social cognitive costs and perceived behavioral control deficits, while strengthening their value recognition of the services [51].

System quality constitutes the core technical support for the implementation of human-AI collaborative consultation services in OHCs [57]. Based on the DeLone and McLean information systems success model, system quality is defined as a comprehensive assessment of system stability (SYS), system responsiveness (SYR), and operational convenience (OPC). These dimensions jointly form the technical framework of human-AI collaborative consultation. By optimizing system effectiveness, reducing usage barriers, and ensuring service reliability, they empower the efficient collaboration of other elements in the information ecosystem, ultimately driving the healthy development of PS and the service ecosystem [58-60].

SYR is defined as the subjective perception of patients regarding the feedback speed, timeliness, and clarity of the human-AI collaborative consultation system. SYR serves as the effectiveness hub of technical support, optimizing the timeliness of information transmission. High responsiveness not only enhances patients’ communication effectiveness but also increases their real-time satisfaction with the service experience. SYS refers to the patient’s experience of the reliability, error rate, and compatibility of the human-AI collaborative consultation system’s operation. SYS provides the fundamental carrying capacity of the technical environment, ensuring the continuous and reliable operation of services. Stable system operation can effectively reduce patients’ concerns about technical malfunctions and enhance their trust in the service and willingness to use it. The ease of operation refers to patients’ perception of the interface interaction logic, learning difficulty, and operation effectiveness of the human-AI collaborative consultation system. OPC is the user interface hub of technology landing, which directly affects the user experience and the effectiveness of patients. Convenient operation design can significantly reduce the learning cost of patients and enhance their acceptance and satisfaction with the system [61]. During the human-AI collaborative consultation process in OHC, high responsiveness ensures smooth communication, high stability guarantees continuous reliability, and high convenience enhances usage effectiveness. When patients perceive that the system offers fast response speed, operates stably without faults, and features a friendly interface, their satisfaction with the human-AI collaborative consultation service will be significantly enhanced.

Based on the theory of information ecology and integrating the characteristics of human-AI collaborative consultation in OHCs as well as related research, this paper establishes a theoretical model of the influencing factors of PS under human-AI collaborative consultation in OHCs, as shown in Figure 1.

Data processing and analysis were conducted using Amos (version 26.0; IBM Corp) and SPSS (version 26.0; IBM Corp). The reliability of the scale was evaluated using the Cronbach α coefficient and composite reliability (CR). The results indicated that the Cronbach α coefficient and CR values of each variable were all >0.8, indicating that the scale had good internal consistency (Table 3). Validity analysis encompassed content validity, construct validity, convergent validity, and discriminant validity. In terms of content validity, the questionnaire scale was designed based on mature scales and was optimized through translation-backtranslation procedures and a pilot study, thus demonstrating relatively high content validity. Construct validity was tested by confirmatory factor analysis (CFA), which yielded a Kaiser-Meyer-Olkin test value of 0.970 and a significant Bartlett sphericity test result (P<.001), indicating that the data were suitable for factor analysis (Table 4). The CFA results showed the following values: χ²₉₄₃=1.264, root-mean-square error of approximation=0.03, goodness-of-fit index=0.86, normed fit index=0.915, comparative fit index=0.981, and Tucker-Lewis index=0.978, suggesting a good model fit (Table 5). Regarding convergent validity, the factor loadings of each item are all >0.5, the average variance extracted values were all >0.5, and the CR values were all >0.8 (Table 6). The discriminant validity test revealed that the square roots of average variance extracted for each variable were all greater than their correlations with other variables (Table 4), indicating that the scale demonstrated good discriminant validity.

As all the questionnaire data obtained were filled out by the participants within the same time period, there may be homologous bias. To verify whether there is a common method bias in the data, the single-factor CFA method was adopted. All measurement items were loaded onto a single latent factor to test the degree of model fitting. If the model fits well, it indicates that there may be significant common method bias. According to Table 7, all the fitting indicators of the original model (χ²₉₄₃=1.264, root-mean-square error of approximation=0.03, comparative fit index=0.981, Tucker-Lewis index=0.978) are significantly better than those of the single-factor model. The fitting results of the single-factor model are poor, indicating that there is no serious common method bias problem in the data.

To meet the Boolean logic requirements for qualitative comparative analysis, variables must be transformed into sets, and cases must be assigned to these sets before conducting fsQCA—a process known as data calibration. In this process, the values of variables must be constrained to the [0,1] interval (where 0 indicates full nonmembership and 1 indicates full membership). Given that a Likert five-point scale was adopted, the questionnaire data required calibration. First, the mean values of each measurement term are taken as variable reflection values. Considering the relationship between the calibration anchor points and the actual distribution of the samples [71], the direct calibration method is adopted, and the 0.95, 0.5, and 0.05 quantiles are respectively taken as complete membership points, intersection points, and complete nonmembership points [72]. To avoid the situation where the membership fraction is 0.5, subtract 0.001 from the calibrated value of 0.5 [73]. The entire procedure was executed with the calibrate (x, n1, n2, n3) routine in fsQCA 4.1.

In accordance with the fsQCA method, before conducting the configurational analysis, the first step is to perform a necessity analysis of the individual condition variables, with the results expressed through consistency and coverage. Consistency represents the degree to which the condition variables are a subset of the outcome variables. Identifying a necessary condition generally requires a consistency score higher than 0.9 [76]. Coverage represents the extent to which the condition variables explain the outcome. Coverage is meaningful only for conditions that satisfy the consistency requirement, and no generally accepted threshold exists.

Necessity analysis of the calibrated data was performed using fsQCA 4.1 software. The analysis results show that the consistency scores of all condition variables have not reached the necessary standard of 0.9, indicating that no single condition variable independently constitutes a necessary condition for PS. Therefore, further exploration of how combinations of condition variables affect PS is required through configurational path analysis. Table 8 presents the results of the necessity test for individual condition variables.

After completing the univariate necessity analysis, the fsQCA 4.1 software was further used for configuration analysis to explore the sufficiency effect of variable combinations on PS. Referring to the configuration analysis method of Ragin [71], the consistency threshold was set to 0.8, the proportional reduction in inconsistency value to 0.7, and the case threshold to 2. A truth table was constructed, and the configuration solution was generated. The analysis results show that the consistency of all five configurations is higher than 0.9, and the overall consistency is 0.962, significantly higher than the threshold value of 0.75, indicating that these configurations constitute sufficient conditions. Meanwhile, the overall coverage rate was 0.544, indicating that the research results could explain more than half of the cases in the sample. By distinguishing the core conditions that occur simultaneously in the intermediate solution and the parsimony solution from the auxiliary conditions that only appear in the intermediate solution, the configurations with the same core conditions are classified, and ultimately, five types of antecedent configuration patterns that trigger high PS are obtained (Table 9).

From a single-factor perspective, in the configuration leading to high PS, the following conditions are all core or important: high SYR, high IUS, high PSE, high PSA, and high PSEF. High IA, high IUD, and high LRP are auxiliary conditions. It is worth noting that neither nonhigh PU nor nonhigh SRs appear in the configuration. From the perspective of factor combination, a comparison between configurations 3 and 4 reveals that, given the simultaneous presence of SYR, SYS, OPC, IUS, IA, IUD, PSE, PSA, PSEF, LRP, and PEN, PU and SRs exert a substitution effect in enhancing PS.

Based on configuration characteristics and driving logic, five realization paths for high PS were identified, and the generation mechanisms of PS were revealed from the dimensions of technology, information, service, and environment, respectively.

Regarding the technology-information human-driven type guided by SRs (configuration 1), the core antecedent conditions of configuration 1 include SYR, SYS, PSE, PSA, PSEF, and SRs. SYR and stability provide patients with a favorable technical experience, ensuring that they receive rapid feedback and a continuous and stable service experience during use. This effectively avoids negative emotions caused by lag, delay, or system crash and enhances overall usability and trust. On this basis, the empathy, accuracy, and effectiveness of perceived services, as key manifestations of the “serving people” dimension, address patients’ multidimensional needs in terms of emotional resonance, professional judgment, and response speed. As an external environmental factor, social norms play a guiding and reinforcing role in this configuration. Through social transmission mechanisms such as recommendations from relatives and friends, public evaluation, and media coverage, a positive cognitive atmosphere regarding the human-AI collaborative consultation model is created, which enhances patients’ willingness to adopt and confidence in using the system. Although IA, IUD, and LRP are marginal conditions, they provide patients with a reliable and orderly medical environment through their basic supporting role. Furthermore, the impact of PU in this configuration is relatively weak, indicating that the effective combination of the above core conditions has successfully alleviated patients’ concerns about service uncertainty, achieving a transition from “technology availability” to “service trustworthiness.” This fully demonstrates the sustainable development logic of “technology as the foundation, service as the core, and social empowerment” in human-AI collaborative consultation.

Regarding the technology-information human-driven under the inclusiveness of platform ethics (configuration 2), the core conditions of configuration 2 are identical to those of configuration 1, namely SYR, SYS, PSE, PSA, PSEF, and SRs, indicating that technical reliability and service quality perception play universal and key roles in human-AI collaborative consultation. The difference is that in configuration 2, OPC and PEN are not core conditions; in some cases, they are even absent, reflecting that there exists a “compensation mechanism” and “priority differentiation” in achieving PS. Our findings show that although platform operation is somewhat complex, the system’s rapid response and stability can effectively offset the negative impact of the operational threshold, embodying the compensation logic that “efficient response is a form of user experience optimization.” Furthermore, the absence of marginal ethical norms on the platform indicates that, in specific usage scenarios, patients tend to prioritize immediate and tangible service value over abstract long-term ethical guarantees. This phenomenon is particularly evident among user groups that primarily receive low-risk, high-frequency services such as daily health consultations, minor illness diagnoses, and medication advice. This finding indicates that, in specific user demand scenarios, a good service experience itself can partially substitute for the construction of institutional trust, forming a trust generation mechanism whereby “service temperature offsets the cost of norms.” This path highlights the need for smart medical services to be tailored to specific needs. For different user groups and usage scenarios, while maintaining the basic compliance bottom line, priority should be given to strengthening technical performance and service quality to achieve a dynamic balance between effectiveness and sustainability.

Regarding the technology-environment driven type under the guidance of platform specifications (configuration 3), configuration 3 reveals a satisfaction realization path driven by “operational convenience” and “platform ethical norms,” “platform ethical norms,” highlighting the fundamental role of technology accessibility and system credibility in shaping PS among specific user groups and usage scenarios. The convenience of operation plays a key role as “entry empowerment” in this path. By simplifying the consultation process, it enables older adult patients, individuals with low digital literacy, or those seeking help for acute conditions—who may be less familiar with the consultation system—to easily initiate consultations, upload materials, and so on, thereby significantly reducing the usage barriers. This allows them to initiate consultations, submit health information, and receive feedback smoothly, thus enhancing the accessibility and inclusiveness of services. PEN serves as the core support at the environmental level, providing institutional guarantees regarding data privacy protection, algorithmic fairness, service transparency, and accountability traceability. Although they do not directly participate in the service interaction process, they establish a “trusted foundation” for the entire consultation ecosystem. Furthermore, information factors and service perception factors are in a marginal position in this configuration, indicating that when basic availability and institutional security are not yet fully satisfied, higher-level service experience is unlikely to become the main factor influencing satisfaction. This model is particularly suitable for scenarios such as community medical mini-programs and grassroots health service platforms that offer low-frequency and lightweight services to the general public. Its success lies in winning users’ trust through “simplicity + reliability.” Therefore, this path emphasizes that during the process of expanding smart health care toward inclusiveness and grassroots levels, the platform should prioritize building a “low-threshold and highly trustworthy” infrastructure, taking OPC and ethical norms as the key levers to enhance the satisfaction of vulnerable groups, thereby providing strong support for achieving health equity and sustainable development.

Regarding the all-element collaborative drive type under demand-oriented logic (configuration 4), configuration 4 embodies a “full-chain, high-standard” service logic for high-value and high-stickiness users, and represents the highest-level pathway to achieving the effectiveness of human-AI collaborative consultation. Its core feature lies in the comprehensive coordination and deep integration of the four dimensions—technology, information, service, and environment—forming a closed-loop service system characterized by “stable technology, solid information, warm humanistic atmosphere, and excellent environment.” SYR, stability, and OPC build a technical experience advantage. IUS meets patients’ core health demands. Perceiving service elements strengthens the humanistic care of services. PEN and SRs further optimize the service environment. This path is particularly suitable for in-depth health service scenarios such as chronic disease management, long-term follow-up, and rehabilitation guidance. Its users often have a strong dependence on the platform and possess a high level of health literacy and service expectations. They are concerned not only with the immediate outcomes of a single consultation but also place great importance on service continuity, professional consistency, and trust accumulation. Consequently, any shortcoming in any link may lead to a significant decline in overall satisfaction. Configuration 4 indicates that to meet the comprehensive needs of such users, the platform must abandon the “single-point optimization” mindset and shift toward systematic and refined service design. It should provide comprehensive protection across the entire chain, take deep alignment with user needs as its orientation, and build a sustainable, high-quality service ecosystem, thereby enhancing user loyalty and the platform’s core competitiveness.

Regarding the all-element collaborative driven type made inclusive by technology stability (configuration 5), configuration 5 reveals a specific path that can still achieve high satisfaction through the synergy of other core elements against a background of relatively weakened SYS, reflecting users’ differentiated perception of service priorities at specific development stages. Its core conditions include SYR, OPC, IUS, PSE, PSA, and PSEF, forming an experience advantage centered on “quick response, easy and efficient use, trustworthy content, and empathetic service.” SYS is not a necessary condition in this configuration, indicating that during the early stage of rapid platform expansion or technological iteration, some patients have a relatively high tolerance for occasional system failures or brief service interruptions. Their satisfaction depends more on whether the platform can provide timely, accurate, and empathetic responses at critical points. This phenomenon is particularly evident among young users, individuals with acute mild symptoms, or those seeking information-based consultations. They are more concerned with the effectiveness and quality of problem-solving rather than the absolute stability of the system. Furthermore, although legal and regulatory guarantees and PEN have not emerged as core conditions, they have indirectly enhanced the sense of security in use by reducing patients’ PU, playing an “implicit support” role. It is worth noting that this path has phased characteristics and is suitable for the platform growth stage or market acquisition stage. In the long term, if SYS issues occur frequently, it will significantly weaken patients’ trust in the platform. Therefore, the platform still needs to increase technological investment, address the stability deficiencies, and achieve a transition from “function-driven” to “reliably driven” operations to ensure the sustainable improvement of service effectiveness.