Abstract
Background
Hair loss is a sensitive issue that can lower self-esteem and psychological well-being, causing many individuals to avoid face-to-face consultations. Chatbots offer a private and convenient way to deliver information on hair loss treatments, yet little is known about how individuals perceive and adopt these tools.
Objective
This study aimed to identify the factors—including appearance-related self-esteem—that influence the intention to use chatbots designed to provide hair loss medication information.
Methods
A web-based survey was conducted among 500 adults in South Korea interested in hair loss treatment. A conceptual model combining Technology Acceptance Model (TAM) and Health Belief Model (HBM) was developed, including perceived usefulness and ease of use from TAM, and perceived susceptibility, severity, and barriers from HBM. Structural equation modeling (SEM) was used to analyze the data. Mediation effects were tested using bootstrapping, and moderation effects were assessed by comparing unconstrained and constrained models.
Results
Perceived usefulness (β=0.605) and perceived susceptibility (β=0.194) had significant positive effects on the intention to use the chatbot. Perceived ease of use (β=0.464) positively influenced perceived usefulness and indirectly affected the intention to use the chatbot (β=0.281). Appearance-related self-esteem significantly moderated the relationships, strengthening the effects of perceived usefulness and perceived susceptibility on the intention to use the chatbot (Δχ2=14.395).
Conclusions
This study demonstrates that perceived usefulness and perceived susceptibility are key factors in the adoption of chatbots for hair loss treatment information. Appearance-related self-esteem is a significant moderator, suggesting that personalized digital interventions that reinforce self-image may enhance engagement and acceptance.
Keywords: Technology Acceptance Model, Health Belief Model, appearance-related self-esteem, chatbot, alopecia
Introduction
Social media and online resources have significantly impacted various aspects of our lives, particularly in how individuals seek and use healthcare information to guide their decisions.1,2 Digital health interventions encompass a wide range of tools designed to support patients, including mobile health (mHealth) applications, telemedicine or tele-counseling services, and conversational agents such as chatbots. According to previous studies, 74.9% of respondents owned a smartphone or tablet, among whom 30.7% had installed health and fitness apps. 3 Mobile health applications are widely used to support self-monitoring, medication adherence, and chronic disease management, offering scalability and continuous patient engagement; however, they often rely on user-initiated interaction and may require sustained motivation and digital literacy to be effective. 4 Tele-counseling and telemedicine services, on the other hand, enable direct communication with health care professionals and can improve access to care, particularly for patients facing geographical or logistical barriers, but they are resource-intensive and may not provide immediate, on-demand responses. 5
More recently, conversational agents—including both rule-based chatbots and generative artificial intelligence (AI) systems—have been introduced as an alternative modality for delivering health information. In particular, generative AI–based conversational agents, such as large language model (LLM)–based systems, offer flexible and natural dialogue capabilities, enabling more personalized and adaptive interactions. However, concerns remain regarding their reliability, consistency, and potential for generating inaccurate or unsafe medical information, especially in the absence of standardized evaluation frameworks.6–8 Chatbots facilitate access to health content in various formats—text, audio, images, and videos—thus enhancing usability for individuals with low health literacy. 9 Owing to these strengths, chatbots have been applied across diverse healthcare areas, including emotional stress management, pre-visit symptom screening, nutrition education, and medication adherence support.10–14 These applications demonstrate the potential of chatbots as effective tools for health communication, particularly in delivering medication information and enhancing adherence behaviors.15–17 Their effectiveness is often context-dependent and varies across settings,17,18 making it important to understand their unique adoption mechanisms within the broader digital health ecosystem.
Hair loss, or alopecia, affects up to 50% of men and women during their lifetime, 19 with a greater global burden observed among older adults. 20 Furthermore, hair loss is closely related to self-esteem and is associated with various mental health issues, such as anxiety, depression, and reduced self-worth.21–24 For instance, many women undergoing chemotherapy perceive hair loss as more emotionally distressing than undergoing mastectomy, 25 and nearly half of youth with alopecia areata experience anxiety or depression. 26 Thus, improving self-esteem and mental health in hair loss patients is essential. Hair loss should not be regarded as uniformly uncomplicated, as it may in some cases be associated with underlying medical conditions requiring clinical evaluation. Nevertheless, chatbots may provide a useful supplementary channel for accessing general health information in sensitive conditions such as hair loss, particularly when individuals are reluctant to seek face-to-face care. 27 Given that hair loss management frequently requires pharmacological interventions, ensuring access to accurate medication information is critical. Although healthcare professionals remain the most trusted source, chatbots may provide a viable supplementary channel. 28
Despite the growing public interest in using chatbots for emotionally sensitive conditions like hair loss, there is limited understanding of the factors driving user acceptance and adoption in this context. To this end, theoretical frameworks such as the Technology Acceptance Model (TAM) and the Health Belief Model (HBM) have been widely used to explain users’ behavioral intentions in health technology adoption. TAM explains users’ acceptance of technology based on perceived usefulness and perceived ease of use, which together shape behavioral intention.29,30 Meanwhile, HBM focuses on health-related behaviors, emphasizing perceived susceptibility, perceived severity, perceived benefits, and perceived barriers as key drivers.31,32 While each model captures different aspects of behavior—technological vs. health-related—their integration allows for a more comprehensive understanding of user intention toward digital health tools. 33 Therefore, this study aims to identify key factors influencing the acceptance of hair loss medication chatbots by integrating TAM and HBM.
Psychological constructs such as self-efficacy, 34 trust, 35 privacy concerns, 36 perceived stigma, 27 and eHealth literacy have been widely used to explain digital health adoption. 37 These factors capture users’ confidence in performing behaviors, trust in digital systems, concerns about information security, perceived social judgment, and ability to evaluate electronic health information. However, in appearance-sensitive conditions such as hair loss, users’ responses to digital health tools may also depend on how they evaluate their own physical appearance. Appearance-related self-esteem differs from broader constructs such as body image and self-efficacy because it specifically reflects the evaluative meaning of appearance for one’s self-worth. Thus, examining appearance-related self-esteem as a moderator provides a more context-specific explanation of chatbot acceptance in the hair loss treatment context.
Methods
Study Overview and Ethical Consideration
This study aims to identify the factors influencing the intention to use chatbots that provide information about hair loss treatment medication. To achieve this, we 1) developed a chatbot for research purposes, 2) conducted a survey among adults interested in medications for hair loss treatment, and 3) analyzed the factors influencing the intention to use the chatbot utilizing a model integrating TAM and HBM.
This study was approved by the Institutional Review Board of Sungkyunkwan University (number: 2023-10-004) on October 5, 2023. We provided information outlining the study purpose, procedures, guarantees of anonymity and confidentiality, and instructions for accessing the chatbot. Only individuals who voluntarily agreed to participate after fully understanding the study were included. They were informed that they could withdraw at any time. No personally identifiable information was collected, and all data were used solely for research purposes. Prior to participation, all respondents were required to provide electronic informed consent after reviewing the study information presented on the survey platform. In this survey, accessing the chatbot did not require membership registration or the provision of personal information.
Development of the Research Chatbot
The chatbot used in this study was designed as an advice-based informational tool, providing structured content on various aspects of hair loss treatment medications. These included dosage and usage instructions, mechanisms of action, expected efficacy, potential side effects, drug interactions, precautionary information, and coverage under the national health insurance system. To define the relevant scope of questions, we analyzed a dataset of 10,274 posts collected over a five-year period (June 2017 to June 2022) from Daedamo, the largest online alopecia community in South Korea, where patients and the general public actively share experiences and information on alopecia treatments, hair transplantation, and daily management. In addition, clinical practice guidelines and frequently asked questions (FAQs) related to alopecia were analyzed. These FAQs were sourced from academic societies, general hospitals, and specialized alopecia clinics in tertiary hospitals. To ensure accuracy and credibility, responses were developed by referencing authoritative knowledge sources, including textbooks, research articles, official institutional publications, and websites.
The chatbot was built using the open-source platform Kakao i Open Builder and was designed to be accessed through KakaoTalk, South Korea’s most widely used messaging app, available on both mobile and desktop platforms. 38 The user interface was structured to support intuitive navigation, allowing users to scroll vertically or horizontally, select from menus, or press buttons to explore content. To maintain consistency and accuracy in responses, a pre-analyzed set of FAQs and their corresponding response scenarios were integrated into the system, minimizing the risk of providing inconsistent or incorrect answers.
Survey Instruments
This study used a survey to identify the factors influencing the intention to use chatbots for hair loss treatment medications, using an integrated framework of the TAM and HBM. Survey items were adapted from validated instruments used in prior studies, including those applied in digital health and chatbot adoption contexts. The survey collected demographic data (e.g., gender, age, level of education, job status), factors related to TAM and HBM, and respondents’ perceptions of hair loss and appearance-related self-esteem. The survey items were adapted from measurement tools used in previous studies and modified to suit the context of this study.29,39–42 Minor wording adjustments were made to ensure cultural and contextual appropriateness for a Korean audience, particularly regarding the sensitive nature of alopecia. The adaptation process included translation, expert review, and pilot testing to maintain reliability and validity. For instance, ‘perceived usefulness’, a key construct of TAM, was defined as the degree to which a person believes that using a chatbot to provide information on medicines for hair loss treatment will improve their convenience. We measured it through the following four questions: 1-1) Using the chatbot would allow me to obtain the information I want; 1-2) Using the chatbot would enable me to obtain information more quickly; 1-3) The chatbot is a useful tool; 1-4) Using the chatbot helps in the proper use of medications. Additionally, perceived ease of use was measured with questions 1-5 through 1-8, perceived susceptibility with questions 2-1 through 2-4, perceived severity with questions 2-5 through 2-8, perceived barriers with questions 1-9 through 1-12, perceived self-esteem related to appearance with questions 3-1 through 3-6, and intention to use with questions 1-13 through 1-16. A 5-point Likert scale was used for responses, ranging from “strongly disagree” (1) to “strongly agree” (5).
The measurement items used in this study were adapted from validated instruments reported in the previous research to ensure theoretical rigor and cultural appropriateness. Specifically, perceived usefulness and perceived ease of use were adapted from Davis 29 and Venkatesh & Davis, 43 reflecting the core constructs of the TAM. Perceived susceptibility, perceived severity, and perceived barriers were drawn from HBM-based instruments developed by Janz & Becker 40 and Champion, 31 which have been widely validated in health behavior and digital health research. Intention to use was measured using items developed by Venkatesh et al., 42 and appearance-related self-esteem was adapted from Heatherton & Polivy to capture appearance-oriented self-evaluation. 39
The items were contextually modified to reflect the technological characteristics of chatbots and linguistically adjusted for Korean participants to ensure clarity and cultural resonance. In addition, because technological domains differ in their characteristics and users’ perceptions of emerging technologies may evolve over time, the items were carefully refined to accommodate the specific context of health information chatbots and the sociocultural conditions under which the present study was conducted.
Prior to conducting the main survey, a pilot test was performed to assess the clarity, relevance, and contextual appropriateness of the adapted measurement items. A total of 13 participants—including both pharmacists and non-pharmacists—were recruited through face-to-face and online channels. Participants were asked to review each item for comprehensibility and contextual fit to the chatbot setting. In addition, minor wording adjustments were made to improve readability and ensure that the items reflected the context of a chatbot providing alopecia medication information. This iterative process allowed us to refine the survey instrument, enhance content validity, and increase the likelihood of accurate and meaningful responses in the main study. The final survey items for each construct are detailed in Supplemental Material 1.
Sampling Frame and Recruitment
The web-based survey was conducted exclusively online from October 10 to 12, 2023 through Macromill Embrain, a professional survey research agency in the Republic of Korea. The sample size for this study was determined based on Cochran’s formula (Z = 1.96, P = .500, E = 0.05), resulting in a minimum required sample of 385 participants. 44 Considering the finite nature of the population, a finite population correction was applied. Additionally, a power analysis using G*Power (80% power, α = 0.05, medium effect size) was conducted to ensure the adequacy of the sample size for comparative analyses. 45 The participants were adults aged 20 to 69 years, residing in Seoul, South Korea, who were interested in hair loss treatment medications. A quota sampling method was used, ensuring that the sample reflected the composition of the population of hair loss patients, based on the 2021 data from the Korea Health Insurance Review & Assessment Service. 46 The gender ratio was set at 50:50, as was the age ratio for those aged 39 and below versus those aged 40 and above. This reflects patterns of health care utilization across sex and age groups rather than the true prevalence of hair loss in the general population. Interest in hair loss treatment medications was determined by responses to the question, “How interested are you in hair loss treatment medications?” on a 5-point Likert scale. Those who scored 3 or higher were included, while those scoring 1 or 2 were excluded in the survey. Before starting the survey, participants were required to interact with a chatbot that provided information on various aspects of hair loss treatment medications for at least five minutes.
Statistical Analysis
Descriptive frequency analysis was conducted to examine the sociodemographic profile of the survey participants, such as gender, age group, education level, and prior experience or interest in hair loss treatment. Before hypothesis testing, the reliability and validity of the survey items were assessed. Reliability was assessed using Cronbach’s alpha (α), with a threshold of 0.70 or higher considered acceptable based on conventional standards. 47 Confirmatory factor analysis (CFA) was conducted to assess whether the latent variables were well-explained by the observed variables, using maximum likelihood for parameter estimation. Convergent validity was tested using construct reliability (CR) and average variance extracted (AVE). All items met the thresholds of CR ≥ 0.7 and AVE ≥ 0.5, indicating satisfactory convergent validity. Discriminant validity was tested using the AVE values. Discriminant validity is considered acceptable if the square of the correlation between the latent variables is less than the AVE value (AVE > squared correlation). When the squared correlation is generally less than 0.4, discriminant validity is confirmed. 48 A structural equation model (SEM) was used to assess the goodness-of-fit of the integrated TAM and HBM model, and path coefficient analysis was performed. Mediation effects were tested using bootstrapping, and indirect effects were considered significant if the p-value was less than 0.05 or the confidence interval did not include 0.
To examine potential differences in structural relationships across groups, additional analyses were conducted based on gender and experience of taking hair loss treatment medications. Gender differences were evaluated using CR comparisons of path coefficients, while subgroup differences based on experience of taking hair loss treatment medications were assessed using chi-square difference tests between constrained and unconstrained models. The moderating effect of appearance-related self-esteem was examined using multi-group structural equation modeling. Participants were divided into high and low groups based on the median split of appearance-related self-esteem scores. To ensure comparability across groups, measurement invariance was sequentially tested at the configural, metric, and scalar levels. Model comparisons were conducted using chi-square difference tests (Δχ2) as well as supplementary fit indices, with ΔCFI ≤ 0.01 and ΔRMSEA ≤ 0.015 as thresholds for invariance.49,50 After establishing invariance, structural path coefficients were constrained and compared across groups. Significant chi-square differences and changes in path coefficients (Δβ) were interpreted as evidence of moderation.
An exploratory, post hoc age-stratified multigroup structural equation modeling analysis was conducted to compare structural path coefficients between participants aged 20–39 years and those aged ≥40 years (n=250 in each group). Measurement invariance was evaluated at the configural and metric levels before comparing structural paths across age groups. After metric invariance was established, a model constraining all 6 structural paths simultaneously to equality was compared with the metric-invariance model, followed by 1-df equality-constraint tests for each structural path. Holm adjustment was applied to the 6 path-specific between-group comparisons. Descriptive between-group comparisons were also conducted for prior chatbot use, perceived ease of use, and intention to use. Data were coded for statistical analysis, which was conducted using SPSS 27.0 and AMOS 27.0.
Results
Participant Characteristics
The survey was conducted exclusively online through a professional survey agency from October 10 to 12, 2023. A total of 4,673 invitations were distributed, and responses from individuals who were not interested in hair loss treatment medications or who provided insincere answers were excluded. The survey continued until 500 completed and valid responses were obtained, yielding a response rate of 10.7% (500/4673). Table 1 summarizes the demographic characteristics of participants. Gender was evenly distributed with 250 males (50.0%) and 250 females (50.0%). Age was also balanced, with half of the participants aged 39 and below and the other half aged 40 and above. Most respondents held a college degree or higher (89.2%), and the majority were employed in office or management-related occupations (59.8%).
Table 1.
Characteristics of survey participants.
| Characteristics | Values, n (%) |
|---|---|
| Gender | |
| Male | 250(50.0) |
| Female | 250(50.0) |
| Age (years) | |
| 20-29 | 74(14.8) |
| 30-39 | 176(35.2) |
| 40-49 | 120(24.0) |
| 50-59 | 98(19.6) |
| >60 | 32(6.4) |
| Level of education | |
| Less than university education | 54(10.8) |
| Bachelor’s degree | 376(75.2) |
| Master’s degree (and above) | 70(14.0) |
| Job status | |
| Office job | 299(59.8) |
| Profession | 50(10.0) |
| Sales and service | 29(5.8) |
| Self-employed | 27(5.4) |
| Housewives | 38(7.6) |
| Students | 22(4.4) |
| Others | 35(7.0) |
| Frequency of mobile chatbot usage | |
| Never use | 52(10.4) |
| Rarely use | 90(18.0) |
| Occasionally use | 203(40.6) |
| Frequently use | 117(23.4) |
| Very frequently use | 38(7.6) |
| Hair loss-related status | |
| Experience of hair loss that led to considering taking medication or using products | 276(55.2) |
| Family history of hair loss | 274(54.8) |
| Experience of taking hair loss treatment medications | 131(26.2) |
Results of SEM of competitive models for main hypothesis testing
The goodness-of-fit analysis of the research model indicated satisfactory levels. For the discriminant validity analysis, the square root of each latent variable’s AVE exceeded its correlations with other latent variables. Additionally, each observed variable had the highest cross-loading with its corresponding latent variable compared to other latent variables in the model. Moreover, all constructs demonstrated high internal consistency, with Cronbach’s alpha values above .841, indicating excellent measurement reliability even at the lowest observed level. Detailed information on this can be found in Supplemental Material 2 and 3.
SEM revealed that perceived usefulness (β=0.605, P <.001) and perceived susceptibility (β = 0.194, P <.001) had significant positive effects on the intention to use the chatbot (Table 2). This indicates that the more useful the chatbot was perceived to be, and the more susceptible participants felt to hair loss, the higher their intention to use it. Perceived ease of use (β=0.464, P <.001) had a significant positive effect on perceived usefulness, suggesting that the easier the chatbot was to use, the more useful it was perceived to be.
Table 2.
Results of structural equation modeling for factors influencing the intention to use the chatbot.
| Structural path | β | S.E. | C.R. | P |
|---|---|---|---|---|
| Perceived ease of use → Perceived usefulness | 0.464 | 0.047 | 9.424 | <.001 |
| Perceived usefulness → Intention to use | 0.605 | 0.055 | 11.122 | <.001 |
| Perceived ease of use → Intention to use | 0.000 | 0.055 | -0.002 | .998 |
| Perceived barriers → Intention to use | 0.063 | 0.042 | 1.288 | .198 |
| Perceived susceptibility → Intention to use | 0.194 | 0.061 | 4.166 | <.001 |
| Perceived severity → Intention to use | -0.024 | 0.050 | -0.527 | .598 |
Abbreviation: SE = standard error; C.R. = Critical Ratio.
Level of goodness-of-fit: χ2(df=240) = 582.619 (p<0.001), Goodness-of-Fit Index= 0.909, Adjusted Goodness-of-Fit Index= 0.886, Normed Fit Index= 0.923, Tucker-Lewis index= 0.946, Comparative fit index = 0.953, Root Mean Square Residual = 0.045, Root Mean Square Error of Approximation = 0.053.
Perceived ease of use, perceived barriers, and perceived severity did not show significant direct effects on the intention to use the chatbot. Therefore, we tested the mediation effect by examining the model fit, path coefficients, and statistical significance of the indirect effects. Perceived ease of use had a statistically significant indirect effect on the intention to use the chatbot, mediated through perceived usefulness (Path: perceived ease of use → perceived usefulness → intention to use the chatbot; β = 0.281, standard error = 0.045, 95% confidence interval =0.191-0.365, P =.001). These results indicate that perceived ease of use did not directly influence the intention to use the chatbot but exerted an indirect effect through perceived usefulness, confirming a full mediation effect. Accordingly, the structural model is presented in Figure 1.
Figure 1.

The structural model of the study.
Additional Subgroup Analyses
Additional subgroup analyses were conducted to examine whether the structural relationships differed by gender and experience of taking hair loss treatment medications. For gender, the effect of perceived usefulness on intention to use was significant in both male and female groups, but critical ratio comparisons showed no statistically significant differences in path coefficients between groups. Similarly, multigroup analyses showed no statistically significant differences in structural paths between participants with and without experience of taking hair loss treatment medications. A marginal difference was observed only for the path from perceived susceptibility to intention to use (Δχ2 = 3.361, df = 1, P = .067). Detailed results are provided in Supplemental Material 4.
An exploratory age-stratified multigroup structural equation modeling analysis compared participants aged 20–39 years with those aged ≥40 years (n=250 in each group). Metric invariance was supported across the 2 age groups (Δχ2[18]=25.034, P=.124). Constraining all 6 structural paths simultaneously to equality significantly worsened model fit relative to the metric-invariance model (Δχ2[6]=15.628, P=.016), indicating age-related differences in the structural relationships. Follow-up equality-constraint tests showed that the path from perceived usefulness to intention to use differed significantly between the age groups (Δχ2[1]=7.846, P=.005; Holm-adjusted P=.031). The standardized path coefficient was larger among participants aged 20–39 years than among those aged ≥40 years (β=.714 vs β=.511). No other between-group differences in the structural paths remained significant after correction for multiple comparisons (Supplemental Table 1 – 5). Participants aged 20–39 years reported more frequent prior chatbot use (mean 3.19, SD 1.08 vs mean 2.80, SD 1.01; P<.001) and higher perceived ease of use (mean 4.04, SD 0.74 vs mean 3.83, SD 0.69; P=.001) than participants aged ≥40 years. In contrast, mean intention to use did not differ between the groups (mean 3.15, SD 0.90 vs mean 3.16, SD 0.75; P=.924).
Moderating Effect of Appearance-Related Self-Esteem
We tested the moderating role of appearance self-esteem using multi-group SEM (high vs. low). Chi-square difference tests indicated significant group differences for two structural paths: H6-1 (Δχ2 = 3.902, df = 1, p = .048) and H6-3 (Δχ2 = 4.752, df = 1, p = .029). The remaining hypothesized moderation paths (H6-2, H6-4, H6-5, H6-6) were not significant (all p > .05). The overall comparison across all constrained paths was significant (Δχ2 = 19.328, df = 6, p = .004), suggesting that ASE partially moderates the structural model. Incremental fit index changes were minimal (ΔNFI ≤ .002; ΔIFI ≤ .003; ΔRFI ≤ .001; ΔTLI ≤ .001), indicating that moderation effects are localized to specific paths rather than broadly altering model fit (Table 3).
Table 3.
Multi-group moderation analysis results.
| Path Constraint | df | Δχ2 (CMIN) | P | ΔNFI | ΔIFI | ΔRFI | ΔTLI |
|---|---|---|---|---|---|---|---|
| H6-1 (a1 = b1) | 1 | 3.902 | 0.048 | 0.000 | 0.001 | 0.000 | 0.000 |
| H6-2 (a2 = b2) | 1 | 0.150 | 0.699 | 0.000 | 0.000 | 0.000 | 0.000 |
| H6-3 (a3 = b3) | 1 | 4.752 | 0.029 | 0.001 | 0.001 | 0.000 | 0.000 |
| H6-4 (a4 = b4) | 1 | 1.806 | 0.179 | 0.000 | 0.000 | 0.000 | 0.000 |
| H6-5 (a5 = b5) | 1 | 2.124 | 0.145 | 0.000 | 0.000 | 0.000 | 0.000 |
| H6-6 (a6 = b6) | 1 | 0.434 | 0.510 | 0.000 | 0.000 | 0.000 | 0.000 |
| Overall model | 6 | 19.328 | 0.004 | 0.002 | 0.003 | 0.001 | 0.001 |
Δχ2 = chi-square difference test; df = degrees of freedom; NFI = Normed Fit Index; IFI = Incremental Fit Index; RFI = Relative Fit Index; TLI = Tucker–Lewis Index. Significant moderation effects are highlighted (p < .05).
Participants with higher appearance-related self-esteem showed strengthened effects of perceived usefulness and perceived susceptibility on the intention to use the chatbot. No significant moderation effects were found for the other hypothesized paths (Table 4).
Table 4.
Moderation effect of appearance-related self-esteem.
| Structural path | Low appearance-related self-esteem (N=240) | High appearance-related self-esteem (N=260) | ||||||
|---|---|---|---|---|---|---|---|---|
| β | C.R. | P | β | C.R. | P | |||
| Perceived usefulness | → | Intention to use | 0.539 | 6.609 | <.001 | 0.646 | 9.031 | <.001 |
| Perceived ease of use | → | Intention to use | -0.031 | -0.317 | .752 | -0.059 | -0.768 | .443 |
| Perceived susceptibility | → | Intention to use | 0.188 | 2.378 | .017 | 0.266 | 4.254 | <.001 |
| Perceived severity | → | Intention to use | -0.126 | -1.633 | .102 | 0.082 | 1.399 | .162 |
| Perceived barriers | → | Intention to use | -0.087 | -1.023 | .306 | 0.070 | 1.114 | .265 |
| Perceived ease of use | → | Perceived usefulness | 0.480 | 6.859 | <.001 | 0.438 | 6.231 | <.001 |
| Δχ2=14.395, Δdf=5, p=.013 | ||||||||
Abbreviation: C.R. = Critical Ratio.
Level of goodness-of-fit: χ2=878.410, degree of freedom=480, P=.000, Normed Fit Index=0.889, Incremental fit index = 0.946, Tucker-Lewis index = 0.938, Comparative fit index = 0.946, Root Mean Square Error of Approximation = 0.041.
Discussion
This study aimed to understand the acceptance of chatbots providing information about hair loss treatment medications. The findings demonstrated that perceived usefulness and perceived susceptibility positively influenced the intention to use the chatbot. Ease of use alone does not directly determine the intention to use a chatbot, but it influences perceptions of the chatbot’s usefulness. Additional subgroup analyses based on gender and experience of taking hair loss treatment medications did not reveal statistically significant differences in the structural relationships, suggesting that the proposed model was broadly stable across these subgroups.
The exploratory age-stratified analysis provided additional evidence regarding potential subgroup differences. Although mean intention to use did not differ between the age groups, the association between perceived usefulness and intention to use was stronger among participants aged 20–39 years than among those aged ≥40 years. Younger participants also reported more frequent prior chatbot use and greater perceived ease of use, consistent with greater familiarity with digital technologies. These findings suggest that age group may be associated with differences in how perceived usefulness translates into behavioral intention rather than the overall level of intention itself. However, because the analysis was post hoc and only one structural-path difference remained significant after correction for multiple comparisons, the findings should be interpreted cautiously and confirmed in future studies.
The moderating effect of appearance-related self-esteem strengthened the positive effects of perceived usefulness and perceived susceptibility on the intention to use the chatbot. This suggests that individuals with higher appearance-related self-esteem may be more likely to translate their perceptions of usefulness and susceptibility into actual behavioral intentions. This finding aligns with previous literature highlighting the role of self-esteem in digital health technology adoption, particularly in conditions affecting personal appearance (e.g., alopecia, skin disorders). Beyond its role as a moderating factor, appearance-related self-esteem may also inform the adaptation of communication strategies to users’ psychological characteristics, potentially improving engagement with digital health tools. While previous studies have primarily focused on constructs such as self-efficacy,51,52 this study introduces appearance-related self-esteem as a context-specific moderating factor in chatbot acceptance for an appearance-sensitive condition. This extends existing technology acceptance research by highlighting the potential importance of self-perception and personal image in shaping responses to digital health tools. 53
By identifying appearance-related self-esteem as a moderating variable, this study offers practical insights into the design and development of digital health tools, particularly for conditions involving changes in physical appearance. Future research could explore the moderating role of appearance-related self-esteem in various health conditions (e.g., skin disorders, obesity) and health tools (e.g., counseling apps, self-management apps) to better understand its influence on the acceptance of digital health technologies. Such an approach could serve as an effective strategy for enhancing the efficacy of non-face-to-face communication tools, like chatbots, in managing conditions that affect appearance and identity.
Our finding that perceived ease of use did not directly influence behavioral intention, but did significantly enhance perceived usefulness, is consistent with prior evidence showing that ease of use primarily operates through usefulness in driving technology acceptance. 54 Most Koreans are already highly adept at using smartphones and messaging apps, which may explain the similar results in this study. Similarly, perceived barriers did not significantly affect the intention to use the chatbot, likely because participants were already familiar with using messaging apps and faced few obstacles, especially since the chatbot was free to use. Additionally, perceived severity did not significantly influence the intention to use the chatbot. This is consistent with studies on health behaviors, where perceived susceptibility plays a stronger role than perceived severity once a certain threshold of severity awareness is reached,40,55–57 which likely applies to the condition of hair loss in this study as well.
This study has several limitations that should be considered when interpreting the findings. First, this study faced certain limitations inherent to its design, many of which align with common challenges encountered in digital health research. As our primary goal was to identify the key factors influencing the acceptance and use of chatbots that provide information about hair loss medications, we employed a controlled, scenario-based design to ensure internal validity and isolate user responses to specific attributes. While this approach enhanced the consistency, reliability, and interpretability of the findings, it inevitably limited the ecological validity of user–chatbot interactions, as it could not fully capture the dynamic and adaptive nature of real-world conversational experiences. Such rule-based, scenario-driven chatbots offer advantages in research precision and user experience control but lack the flexibility and personalization characteristic of modern AI-driven systems. In particular, the system was not able to accommodate complex or unanticipated user queries beyond the predefined scenarios, which may differ from real-world interactions with more advanced conversational agents. In addition, compared with generative AI–based chatbots, rule-based systems may offer a less natural and adaptive user experience, which could influence user perceptions of engagement and usefulness. Recent advances in large language model–based chatbots enable more natural and adaptive interactions, but concerns remain regarding their accuracy, safety, and lack of standardized evaluation in healthcare settings.58,59 Therefore, the findings of this study should be interpreted in the context of structured, rule-based systems rather than fully generalizable to all AI chatbot technologies. Future research should explore hybrid or generative AI–driven chatbots, such as those based on LLMs, which can support more natural, adaptive, and personalized health communication. Furthermore, comparative investigations between rule-based and generative chatbot architectures may provide deeper insight into how different interaction styles influence user trust, engagement, and long-term behavioral outcomes.
Second, a limitation concerns the generalizability of the findings. Since this study focused on chatbots for hair loss treatment information, it may be difficult to directly apply the results to other health issues or demographic groups. Future research should replicate this study across various health contexts and diverse populations to validate the universal applicability of the integrated TAM and HBM model. Cross-validation in diverse healthcare contexts—such as chronic disease management or mental health support—would enhance the robustness and generalizability of the findings across different conditions. At the same time, the results are context-specific to hair loss treatment in South Korea, and their applicability may be limited in other settings. Cultural norms, healthcare system characteristics, and societal attitudes toward appearance could shape technology acceptance differently, highlighting the need for cross-cultural research.
Third, there is a possibility of unmeasured confounders that were not accounted for in this study. Prior studies have highlighted that trust in the system,60,61 privacy concerns regarding data security, 62 and cultural factors such as language, norms, and health care system structures may play a critical role in shaping users’ intention to adopt health chatbots. 63 Future research should incorporate these psychological and contextual variables to provide a more comprehensive framework and enhance the ecological validity and cross-cultural generalizability of the findings.
To the best of our knowledge, this study is among the first to propose a theory-driven quantitative framework for understanding chatbot acceptance in the context of hair loss. Although therapeutic bonds have long been regarded as the unique domain of human-delivered psychotherapy, accumulating evidence suggests that digital therapeutics, such as mental health chatbots, are likewise capable of establishing meaningful alliances with users.64,65 Chatbots can elicit positive psychological outcomes by delivering positive psychology interventions in an accessible format, thereby reducing anxiety and psychosocial distress in individuals facing health challenges. 66 Given our findings, which underscore the moderating role of appearance-related self-esteem, we suggest that chatbots for hair loss treatment may be more effective when they include features that psychologically support users’ appearance-related concerns. Such features might include empowering language, success stories from similar users, framing that normalizes hair loss, or links to mental health support resources. These elements, while not directly tested in this study, are theoretically aligned with our moderation findings and may enhance both engagement and adherence in real-world chatbot implementations. Future work should empirically test these features to optimize the design of AI-driven digital health tools for appearance-sensitive conditions.
Additionally, to enhance the generalizability and contextual robustness of our framework, future research should include replication studies across different racial, cultural, and demographic groups. Explore cultural differences in the role of appearance-related self-esteem, as perceptions of appearance and related stigma can vary significantly across societies. Investigate potential interactions between appearance-related self-esteem and other psychological variables, such as self-efficacy, health anxiety, or social support, in influencing chatbot adoption. Comparative studies on different types of chatbots—considering factors such as functional sophistication, content quality, and interface design—are also warranted. Compare chatbot designs, such as AI-driven conversational agents versus structured, menu-based designs, to assess their relative influence on perceived usefulness, ease of use, and adoption intention. Furthermore, expanding this framework to a broader range of health conditions could help identify condition-specific patterns in chatbot acceptance and usage behavior. Validate the integrated TAM-HBM model across diverse medical conditions and populations, particularly those involving visible appearance changes (e.g., dermatological conditions, obesity). Addressing these areas could improve the reliability and applicability of digital health acceptance models, ultimately enhancing the effectiveness and accessibility of chatbot-based interventions in diverse medical contexts.
At the same time, it is important to recognize that chatbots should be understood as supplementary informational tools rather than substitutes for clinical diagnosis or individualized treatment planning.62,67 Although chatbot-based systems can improve access to health information, they should not be regarded as substitutes for professional diagnosis or treatment. In the context of disease including hair loss, which may be associated with clinically significant underlying conditions, overreliance on chatbot-provided information may delay appropriate medical evaluation. Therefore, such tools should be positioned as complementary resources that guide users toward informed decision-making and timely consultation with healthcare professionals.
Conclusions
Accurately predicting user acceptance is essential for the effective implementation of digital health technologies. This study validated an integrated model based on the Technology Acceptance Model and the Health Belief Model, demonstrating that perceived usefulness and perceived susceptibility significantly influence the intention to use chatbots for hair loss treatment information. Appearance-related self-esteem moderated these relationships, suggesting that psychological traits tied to personal appearance can meaningfully shape technology adoption. These findings highlight the importance of considering users’ emotional and psychological characteristics—particularly in health contexts where appearance is a central concern. To this end, personalized design strategies that support self-esteem and emotional well-being may enhance both the adoption and sustained use of digital health tools across diverse care settings.
Supplemental material
Supplemental material for Intention to use chatbots for hair loss treatment: A structural equation modeling study examining the moderating role of appearance-related self-esteem by Hyun-Gyu Choi, Sun-Kyeong Park, Kyung Ah Ban, Wonsang Robert Yu and Eui-Kyung Lee in Digital Health.
Acknowledgments
The authors would like to express their gratitude to all individuals who participated in the survey and to colleagues who provided valuable comments during the preparation of this manuscript.
Appendix.
Abbreviations
- AVE
average variance extracted
- CFA
confirmatory factor analysis
- CR
construct reliability
- FAQ
frequently asked questions
- HBM
health belief model
- SEM
structural equation modeling
- TAM
technology acceptance model.
Author contributions: Hyun-Gyu Choi and Sun-Kyeong Park are co–first authors who equally contributed to Conceptualization, Methodology, and Writing – original draft. Eui-Kyung Lee, as corresponding author, supervised the study and contributed to Project administration and Writing – review & editing. Kyung Ah Ban and Wonsang Robert Yu contributed to Data curation, Validation, and Writing – review & editing. All authors approved the final version of the manuscript.
Funding: The authors received no financial support for the research, authorship, and/or publication of this article.
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Statement of generative AI use: Generative AI was employed exclusively for linguistic purposes, including grammar refinement and stylistic clarity. No content generation, data analysis, or interpretation was conducted by AI. The authors carefully reviewed and edited all text and bear sole responsibility for the manuscript’s intellectual content.
Supplemental material: Supplemental material for this article is available online.
ORCID iDs
Sun-Kyeong Park https://orcid.org/0000-0003-4421-4513
Kyung Ah Ban https://orcid.org/0009-0007-9616-3179
Eui-Kyung Lee https://orcid.org/0000-0003-0601-7754
Ethical considerations
This study was approved by the Institutional Review Board of Sungkyunkwan University (SKKU 2023-10-004) on October 5, 2023.
Consent to participate
All participants provided electronic informed consent before participation.
Data Availability Statement
The survey data used in this study are not publicly available due to privacy restrictions but are available from the corresponding author upon reasonable request.*
References
- 1.Hawkins CM, DeLaO AJ, Hung C. Social Media and the Patient Experience. J Am Coll Radiol 2016; 13: 1615–1621. 10.1016/j.jacr.2016.09.006 [DOI] [PubMed] [Google Scholar]
- 2.Song H, Omori K, Kim J, et al. Trusting Social Media as a Source of Health Information: Online Surveys Comparing the United States, Korea, and Hong Kong. J Med Internet Res 2016; 18: e25. 10.2196/jmir.4193 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Makowsky MJ, Jones CA, Davachi S. Prevalence and Predictors of Health-Related Internet and Digital Device Use in a Sample of South Asian Adults in Edmonton, Alberta, Canada: Results From a 2014 Community-Based Survey. JMIR Public Health Surveill 2021; 7: e20671. 10.2196/20671 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Pérez-Jover V, Sala-González M, Guilabert M, et al. Mobile apps for increasing treatment adherence: Systematic Review. J Med Internet Res 2019; 21(6): e12505. 10.2196/12505 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Totten AM, Womack DM, Eden KB, et al. Telehealth: Mapping the Evidence for Patient Outcomes From Systematic Reviews. Agency for Healthcare Research and Quality, 2016. Report No: 16-EHC034-EF. [PubMed] [Google Scholar]
- 6.Laranjo L, Dunn AG, Tong HL, et al. Conversational agents in healthcare: a systematic review. J Am Med Inform Assoc 2018; 25(9): 1248–1258. 10.1093/jamia/ocy072 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Singhal K, Azizi S, Tu T, et al. Large language models encode clinical knowledge. Nature 2023; 620: 172–180. 10.1038/s41586-023-06088-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Ayers JW, Poliak A, Dredze M, et al. Comparing Physician and Artificial Intelligence Chatbot Responses to Patient Questions Posted to a Public Social Media Forum. JAMA Intern Med 2023; 183(6): 589–596. 10.1001/jamainternmed.2023.0580 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Laumer S, Maier C, Gubler FT. Chatbot acceptance in healthcare: Explaining user adoption of conversational agents for disease diagnosis. In: Proceedings of the 27th European Conference on Information Systems (ECIS), Research Papers, paper 88, 2019. https://aisel.aisnet.org/ecis2019_rp/88/ [Google Scholar]
- 10.Crutzen R, Peters GJY, Portugal SD, et al. An artificially intelligent chat agent that answers adolescents' questions related to sex, drugs, and alcohol: an exploratory study. J Adolesc Health 2011; 48: 514–519. 10.1016/j.jadohealth.2010.09.002 [DOI] [PubMed] [Google Scholar]
- 11.Fadhil A. Can a chatbot determine my diet?: Addressing Challenges of Chatbot Application for Meal Recommendation. arXiv:1802.09100 ; 2018. [Google Scholar]
- 12.Ghosh S, Bhatia S, Bhatia A. Quro: Facilitating User Symptom Check Using a Personalised Chatbot-Oriented Dialogue System. Stud Health Technol Inform 2018; 252: 51–56. [PubMed] [Google Scholar]
- 13.Hess GI, Fricker G, Denecke K. Improving and Evaluating eMMA's Communication Skills: A Chatbot for Managing Medication. Stud Health Technol Inform 2019; 259: 101–104. [PubMed] [Google Scholar]
- 14.Park J-J. A development of chatbot for emotional stress recognition and management using NLP. Trans Korean Inst Elect Eng 2018; 67: 954–961. [Google Scholar]
- 15.Ramadhani CN. Chatbots in pharmacy: A boon or a bane for patient care and pharmacy practice? Sciences of Pharmacy 2023; 2: 117–133. 10.58920/sciphar02030001 [DOI] [Google Scholar]
- 16.Wu Y, Zhang J, Ge P, et al. Application of Chatbots to Help Patients Self-Manage Diabetes: Systematic Review and Meta-Analysis. J Med Internet Res 2024; 26: e60380. 10.2196/60380 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Xue J, Zhang B, Zhao Y, et al. Evaluation of the Current State of Chatbots for Digital Health: Scoping Review. J Med Internet Res 2023; 25: e47217. 10.2196/47217 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Tudor Car L, Dhinagaran DA, Kyaw BM, et al. Conversational agents in health care: scoping review and conceptual analysis. J Med Internet Res 2020; 22: e17158. 10.2196/17158 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Mounsey AL, Reed SW. Diagnosing and treating hair loss. Am Fam Physician 2009; 80: 356–362. [PubMed] [Google Scholar]
- 20.Karimkhani C, Dellavalle RP, Coffeng LE, et al. Global Skin Disease Morbidity and Mortality: An Update From the Global Burden of Disease Study 2013. JAMA Dermatol 2017; 153: 406–412. 10.1001/jamadermatol.2016.5538 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Can G, Demir M, Erol O, et al. A comparison of men and women's experiences of chemotherapy-induced alopecia. Eur J Oncol Nurs 2013; 17: 255–260. 10.1016/j.ejon.2012.06.003 [DOI] [PubMed] [Google Scholar]
- 22.Dougherty L. Using nursing diagnoses in prevention and management of chemotherapy-induced alopecia in the cancer patient. Int J Nurs Terminol Classif 2007; 18: 142–149. 10.1111/j.1744-618X.2007.00069.x [DOI] [PubMed] [Google Scholar]
- 23.Rathnayake D, Sinclair R. Male androgenetic alopecia. Expert Opin Pharmacother 2010; 11: 1295–1304. 10.1517/14656561003752730 [DOI] [PubMed] [Google Scholar]
- 24.Sinclair R, Patel M, Dawson TL, Jr., et al. Hair loss in women: medical and cosmetic approaches to increase scalp hair fullness. Br J Dermatol 2011; 165(Suppl 3): 12–18. 10.1111/j.1365-2133.2011.10630.x [DOI] [PubMed] [Google Scholar]
- 25.Browall M, Gaston-Johansson F, Danielson E. Postmenopausal women with breast cancer: their experiences of the chemotherapy treatment period. Cancer Nurs 2006; 29: 34–42. 10.1097/00002820-200601000-00006 [DOI] [PubMed] [Google Scholar]
- 26.Ghanizadeh A, Ayoobzadehshirazi A. A review of psychiatric disorders comorbidities in patients with alopecia areata. Int J Trichology 2014; 6: 2–4. 10.4103/0974-7753.136746 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Miles O, West R, Nadarzynski T. Health chatbots acceptability moderated by perceived stigma and severity: A cross-sectional survey. Digit Health 2021; 7: 20552076211063012. 10.1177/20552076211063012 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Dilles T, Heczkova J, Tziaferi S, et al. Nurses and Pharmaceutical Care: Interprofessional, Evidence-Based Working to Improve Patient Care and Outcomes. Int J Environ Res Public Health 2021; 18(11): 5973. 10.3390/ijerph18115973 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Davis FD. Perceived Usefulness, Perceived Ease of Use and User Acceptance of Information Technology. MIS quarterly 1989; 13(3): 319–340. [Google Scholar]
- 30.Marangunić N, Granić A. Technology acceptance model: a literature review from 1986 to 2013. Universal access in the information society 2015; 14: 81–95. 10.1007/s10209-014-0348-1 [DOI] [Google Scholar]
- 31.Champion VL, Skinner CS. The health belief model. In: Glanz K, Rimer BK and Viswanath K (eds) Health behavior and health education: Theory, research, and practice 4th ed. San Francisco, CA: Jossey-Bass, 2008, pp. 45–65. [Google Scholar]
- 32.Hochbaum GM. Public participation in medical screening programs: A socio-psychological study. US Department of Health, Education, and Welfare, Public Health Service, 1958. [Google Scholar]
- 33.Ahadzadeh AS, Pahlevan Sharif S, Ong FS, et al. Integrating health belief model and technology acceptance model: an investigation of health-related internet use. J Med Internet Res 2015; 17: e45. 10.2196/jmir.3564 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Bandura A. Self-efficacy: toward a unifying theory of behavioral change. Psychol Rev 1977; 84(2): 191–215. 10.1037/0033-295X.84.2.191 [DOI] [PubMed] [Google Scholar]
- 35.Gefen D, Karahanna E, Straub DW. Trust and TAM in online shopping: an integrated model. MIS Q 2003; 27(1): 51–90. 10.2307/30036519 [DOI] [Google Scholar]
- 36.Leschanowsky A, Rech S, Popp B, et al. Evaluating privacy, security, and trust perceptions in conversational AI: A systematic review. Comput Human Behav 2024; 159: 108344. 10.1016/j.chb.2024.108344 [DOI] [Google Scholar]
- 37.Norman CD, Skinner HA. eHEALS: the eHealth Literacy Scale. J Med Internet Res 2006; 8(4): e27. 10.2196/jmir.8.4.e27 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Piao M, Ryu H, Lee H, et al. Use of the Healthy Lifestyle Coaching Chatbot App to Promote Stair-Climbing Habits Among Office Workers: Exploratory Randomized Controlled Trial. JMIR Mhealth Uhealth 2020; 8: e15085. 10.2196/15085 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Heatherton TF, Polivy J. Development and validation of a scale for measuring state self-esteem. Journal of Personality and Social psychology 1991; 60: 895–910. 10.1037/0022-3514.60.6.895 [DOI] [Google Scholar]
- 40.Janz NK, Becker MH. The Health Belief Model: a decade later. Health Educ Q 1984; 11: 1–47. 10.1177/109019818401100101 [DOI] [PubMed] [Google Scholar]
- 41.Strecher VJ, Rosenstock IM. The health belief model. In: Baum A, Newman S, Weinman J, West R and McManus C (eds) Cambridge handbook of psychology, health and medicine. Cambridge: Cambridge University Press, 1997, pp. 113–117. [Google Scholar]
- 42.Venkatesh V, Thong JYL, Xu X. Consumer acceptance and use of information technology: extending the unified theory of acceptance and use of technology. MIS quarterly 2012; 36: 157–178. 10.2307/41410412 [DOI] [Google Scholar]
- 43.Venkatesh V, Davis FD. A model of the antecedents of perceived ease of use: Development and test. Decision sciences 1996; 27: 451–481. 10.1111/j.1540-5915.1996.tb01822.x [DOI] [Google Scholar]
- 44.Cochran WG. Sampling techniques. John Wiley & Sons, 1977. [Google Scholar]
- 45.Faul F, Erdfelder E, Buchner A, et al. Statistical power analyses using G*Power 3.1: tests for correlation and regression analyses. Behav Res Methods 2009; 41: 1149–1160. 10.3758/BRM.41.4.1149 [DOI] [PubMed] [Google Scholar]
- 46.Health Insurance Review & Assessment Service . Healthcare Bigdata Hub. [cited 2023]; . https://opendata.hira.or.kr [Google Scholar]
- 47.Nunnally JC, Bernstein IH. Psychometric Theory. 3rd edn. McGraw-Hill, 1994. [Google Scholar]
- 48.Fornell C, Larcker DF. Evaluating structural equation models with unobservable variables and measurement error. Journal of marketing research 1981; 18: 39–50. 10.1177/002224378101800104 [DOI] [Google Scholar]
- 49.Chen FF. Sensitivity of goodness of fit indexes to lack of measurement invariance. Structural equation modeling: a multidisciplinary journal 2007; 14: 464–504. 10.1080/10705510701301834 [DOI] [Google Scholar]
- 50.Cheung GW, Rensvold RB. Evaluating goodness-of-fit indexes for testing measurement invariance. Structural equation modeling 2002; 9: 233–255. 10.1207/s15328007sem0902_5 [DOI] [Google Scholar]
- 51.Shahzad MF, Xu S, Zahid H. Exploring the impact of generative AI-based technologies on learning performance through self-efficacy, fairness & ethics, creativity, and trust in higher education. Education and Information Technologies 2025; 30: 3691–3716. 10.1007/s10639-024-12949-9 [DOI] [Google Scholar]
- 52.Yildiz Durak H. Conversational agent-based guidance: Examining the effect of chatbot usage frequency and satisfaction on visual design self-efficacy, engagement, satisfaction, and learner autonomy. Education and Information Technologies 2023; 28: 471–488. 10.1007/s10639-022-11149-7 [DOI] [Google Scholar]
- 53.Merino M, Tornero-Aguilera JF, Rubio-Zarapuz A, et al. Body Perceptions and Psychological Well-Being: A Review of the Impact of Social Media and Physical Measurements on Self-Esteem and Mental Health with a Focus on Body Image Satisfaction and Its Relationship with Cultural and Gender Factors. Healthcare (Basel) 2024; 12: 1396. 10.3390/healthcare12141396 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 54.Park YE, Tak YW, Kim I, et al. User Experience and Extended Technology Acceptance Model in Commercial Health Care App Usage Among Patients With Cancer: Mixed Methods Study. J Med Internet Res 2024; 26: e55176. 10.2196/55176 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 55.Carpenter CJ. A meta-analysis of the effectiveness of health belief model variables in predicting behavior. Health communication 2010; 25: 661–669. 10.1080/10410236.2010.521906 [DOI] [PubMed] [Google Scholar]
- 56.Weinstein ND. The precaution adoption process. Health Psychol 1988; 7: 355–386. 10.1037//0278-6133.7.4.355 [DOI] [PubMed] [Google Scholar]
- 57.Harrison JA, Mullen PD, Green LW. A meta-analysis of studies of the Health Belief Model with adults. Health Educ Res 1992; 7: 107–116. 10.1093/her/7.1.107 [DOI] [PubMed] [Google Scholar]
- 58.Hanss K, Sarma KV, Glowinski AL, et al. Assessing the accuracy and reliability of large language models in psychiatry using standardized multiple-choice questions: cross-sectional study. J Med Internet Res 2025; 27: e69910. 10.2196/69910 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 59.Hua Y, Xia W, Bates D, et al. Standardizing and scaffolding health care AI-chatbot evaluation: systematic review. JMIR AI 2025; 4: e69006. 10.2196/69006 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 60.Shahsavar Y, Choudhury A. User Intentions to Use ChatGPT for Self-Diagnosis and Health-Related Purposes: Cross-sectional Survey Study. JMIR Hum Factors 2023; 10: e47564. 10.2196/47564 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 61.Biro J, Linder C, Neyens D. The Effects of a Health Care Chatbot's Complexity and Persona on User Trust, Perceived Usability, and Effectiveness: Mixed Methods Study. JMIR Hum Factors 2023; 10: e41017. 10.2196/41017 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 62.Li J. Security Implications of AI Chatbots in Health Care. J Med Internet Res 2023; 25: e47551. 10.2196/47551 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 63.Kim M, Kim E, Lee H, et al. A Culturally Tailored Artificial Intelligence Chatbot (K-Bot) to Promote Human Papillomavirus Vaccination Among Korean Americans: Development and Usability Study. Asian Pac Isl Nurs J 2025; 9: e71865. 10.2196/71865 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 64.Beatty C, Malik T, Meheli S, et al. Evaluating the Therapeutic Alliance With a Free-Text CBT Conversational Agent (Wysa): A Mixed-Methods Study. Front Digit Health 2022; 4: 847991. 10.3389/fdgth.2022.847991 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 65.Darcy A, Daniels J, Salinger D, et al. Evidence of human-level bonds established with a digital conversational agent: cross-sectional, retrospective observational study. JMIR Formative Research 2021; 5: e27868. 10.2196/27868 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 66.Greer S, Ramo D, Chang YJ, et al. Use of the Chatbot “Vivibot” to Deliver Positive Psychology Skills and Promote Well-Being Among Young People After Cancer Treatment: Randomized Controlled Feasibility Trial. JMIR Mhealth Uhealth 2019; 7(10): e15018. 10.2196/15018 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 67.Laymouna M, Ma Y, Lessard D, et al. Roles, users, benefits, and limitations of chatbots in health care: rapid review. J Med Internet Res 2024; 26: e56930. 10.2196/56930 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Supplemental material for Intention to use chatbots for hair loss treatment: A structural equation modeling study examining the moderating role of appearance-related self-esteem by Hyun-Gyu Choi, Sun-Kyeong Park, Kyung Ah Ban, Wonsang Robert Yu and Eui-Kyung Lee in Digital Health.
Data Availability Statement
The survey data used in this study are not publicly available due to privacy restrictions but are available from the corresponding author upon reasonable request.*
