Accessibility settings

This paper is in the following e-collection/theme issue:

Design and Usability of Consumer Health Tech and Home Monitoring Devices (441) Digital Mental Health Interventions, e-Mental Health and Cyberpsychology (3102) Human Factors and Usability Case Studies (1279) Eating Disorders (133) Formative Evaluation of Digital Health Interventions (4964) Chatbots and Conversational Agents (1140)

Published on in Vol 13 (2026)

Preprints (earlier versions) of this paper are available at https://preprints.jmir.org/preprint/90783, first published .
Woman in glasses looking at her smartphone

Body Image and Eating Disorder Education Chatbot, JEM, in Australia and Canada: First 6-Month Real-World Survey Evaluation

Body Image and Eating Disorder Education Chatbot, JEM, in Australia and Canada: First 6-Month Real-World Survey Evaluation

Authors of this article:

Gemma Sharp1 Author Orcid Image ;   Sara Marini2 Author Orcid Image ;   Emily Tam2 Author Orcid Image ;   Hao Hu1 Author Orcid Image
Article Authors Cited by Tweetations Metrics

    1Department of Neuroscience, Monash University, 99 Commercial Road, Melbourne, Victoria, Australia

    2National Eating Disorder Information Centre, Toronto, ON, Canada

    Corresponding Author:

    Gemma Sharp, PhD


    Background: Body image dissatisfaction, disordered eating, and eating disorders represent significant public health concerns; however, many affected individuals never access evidence-based support. We co-designed and developed a rule-based chatbot, JEM, which conducts conversations addressing evidence-based psychoeducation and psychotherapeutic microinterventions. We previously demonstrated the feasibility, acceptability, and preliminary satisfaction of the JEM chatbot in a research setting. However, broader satisfaction, experiences, and user-reported outcomes in real-world settings have not yet been investigated.

    Objective: This study aims to conduct a real-world evaluation of the JEM chatbot in Australia and Canada, the two countries that have hosted a deployment of the chatbot to date. Specifically, we aim to explore user satisfaction and experiences with the chatbot and within-session differences in user mood and body image satisfaction when completing the chatbot’s microinterventions.

    Methods: Respondents were users of the JEM chatbot aged 13 to 64 years who self-selected to complete a web-based overall evaluation survey (N=230; n=122 in Australia and n=108 in Canada) over a 6-month period. This evaluation survey included user demographic characteristics, satisfaction measures, and the System Usability Scale. Respondents for the within-session pre-post analyses were JEM chatbot users who chose to complete brief web-based surveys immediately before and after completing one of the chatbot’s microinterventions during the same 6-month period. Sample sizes varied across microinterventions, ranging from 75 to 276 respondents overall (Australia: n=34‐146; Canada: n=39‐130). These surveys included validated visual analog scales (VAS) measuring mood (anxiety, depression, happiness, confidence) and body image satisfaction (body size satisfaction, body shape satisfaction, physical attractiveness).

    Results: Demographic characteristics showed that survey respondents were commonly young adult cisgender women and nonbinary individuals across Australia and Canada. Respondent satisfaction with the chatbot was high in both countries (Australia: mean 76.1, SD 22.7; Canada: mean 78.8, SD 14.3), and the usability of the chatbot was rated as “excellent” in both countries (Australia: mean 86.5, SD 16.9; Canada: mean 89.5, SD 11.6) according to the System Usability Scale. Across completed microintervention surveys, patterns of within-session pre-post ratings were broadly similar in Australia and Canada, with effect sizes generally ranging from very small to large across VAS-measured mood and body image outcomes.

    Conclusions: The JEM chatbot achieved high satisfaction and usability ratings. Among respondents who completed pre-post surveys, immediate within-session differences in mood and body image ratings were observed following the completion of chatbot microinterventions. The study findings were broadly similar across Australia and Canada. These results provide evidence of user experience and within-session differences following engagement with JEM and support continued evaluation in future studies.

    JMIR Hum Factors 2026;13:e90783

    doi:10.2196/90783

    Keywords

    eating disorder; body image; mental health; chatbot; conversational agent; artificial intelligence; digital health; psychoeducation; microintervention; evaluation; survey 


    Eating disorders are characterized by disruptions in eating or feeding behaviors that negatively impact an individual’s physical and psychosocial functioning [1]. These disorders are present across ages, genders, and backgrounds, and their prevalence and impact are increasing [2,3]. Eating disorders are complex and involve a combination of biological, psychological, sociocultural, and environmental factors [2]. Body image concerns and body dissatisfaction have a significant association with the development of some eating disorders and extreme weight-loss behaviors [4]. Body image dissatisfaction and eating disorders are major public health concerns contributing to substantial economic and health burdens [5-7]; however, many affected individuals cannot access psychoeducation, early intervention, or timely support.

    In response to the growing demand for accessible mental health support, digital mental health interventions have increased in popularity, including artificial intelligence (AI) chatbots. Chatbots offer advantages such as accessibility, instantaneous responses, and being free of cost to users, and they can act as a pathway to more individualized services [8,9]. They can also be helpful in overcoming barriers such as mental health stigma, which is particularly common in body image distress and eating disorders, or having a lack of services available [8,9]. An expanding body of research is supporting chatbot usage and efficacy in the mental health care sector [10-13]. However, despite the rapid growth in chatbot development, relatively few have been rigorously evaluated in real-world eating disorder or body image–specific contexts, with most evidence derived from controlled or early-stage pilot studies [14]. This limits understanding of how such tools perform under naturalistic conditions, including voluntary use, more diverse user populations, and nonresearch settings. As a result, there remains limited evidence regarding engagement, acceptability, and short-term outcomes of body image– and eating disorder–focused chatbots when deployed at scale.

    To date, a small number of chatbots with eating disorder– and/or body image–focused evidence-based components have been reported in the literature, including Tessa [15], Topity [16], Alex [17], an early-stage unnamed chatbot for adolescents at risk of eating disorders [18], ED ESSI [19,20], and JEM [21] (formerly KIT), which is the focus of the current study. These systems differ in their primary aims and intervention focus, spanning prevention and early intervention (Tessa, Topity), support for individuals on treatment waitlists (ED ESSI), and early-stage or partially evaluated tools (Alex). Only Tessa, Topity, and ED ESSI have demonstrated effectiveness in randomized controlled trials (RCTs), with Tessa reducing weight and shape concerns in young women [15], Topity improving body image outcomes in adolescents [16], and ED ESSI reducing eating disorder pathology and mood-related outcomes in adolescents and adults awaiting treatment [19]. Certain components of Alex have also been evaluated in controlled settings, although the optimized system has not yet been fully tested [22]. Overall, despite promising early findings, there remains a lack of large-scale real-world evaluations of eating disorder and body image–focused chatbots that combine psychoeducation with brief evidence-based skill-building interventions.

    JEM is a rule-based chatbot (ie, operating using predefined conversational pathways rather than generative AI) developed using the Google Dialogflow platform [23]. The chatbot was designed to deliver evidence-based eating disorder–focused psychoeducation alongside brief skills-based microinterventions [21]. The psychoeducation component provides structured information on eating disorders, including prevalence, types, warning signs, health effects, causes, and help-seeking, while the microinterventions focus on brief cognitive behavior therapy (CBT), acceptance and commitment therapy (ACT), and mindfulness-informed exercises targeting cognitive and emotional processes. These microinterventions include education on cognitive distortions or unhelpful thinking styles (CBT), practicing detaching from unhelpful thoughts via cognitive defusion exercises (ACT), and mindful breathing. Both the psychoeducation and microintervention components are delivered using multimodal formats, including text, audio, video, and graphics.

    JEM was co-designed with young people aged 13 years and older with lived experience of body image concerns and/or eating disorders, parents/carers, and health professionals [21]. The co-design process provided iterative feedback on content, tone, and usability, which informed the refinement of both the psychoeducation and microintervention components. Overall, co-design findings supported the acceptability and feasibility of JEM and informed its final structure and delivery format [21]. The efficacy of JEM was not tested in an RCT setting before wider scale deployment within Australia because the development timing coincided with the COVID-19 pandemic when demand for support was high and many health initiatives were decided to be fast-tracked [24,25].

    While it can be argued that RCTs are ideal for establishing evidence of efficacy as they offer structured environments for rigorous hypothesis testing and reduce the influence of external factors, their findings may not always be generalizable [26]. Evaluating a chatbot in a real-world setting is advantageous in this regard. Naturalistic evaluations are beneficial for understanding user experiences, engagement, and outcomes as they offer increased ecological validity by enabling more natural interactions and outcomes [27-29]. Moreover, they can involve a diverse user base, allowing for a better understanding of the chatbot’s usability and effectiveness across various demographics [28,29]. Additionally, real-world evaluations may capture more accurate long-term engagement, as users are not confined by the structured timeframe or artificial environment of an RCT, making it easier to evaluate the chatbot’s sustainability and broader impact [26,27]. However, such designs also introduce limitations including self-selection bias, lack of control conditions, and challenges in causal inference.

    Real-world evaluations of mental health–specific chatbots are seemingly quite rare [30]. For instance, a study of Wysa, a mobile app for mental resilience and well-being, involved 129 voluntary users [31]. Data were collected via anonymous in-app feedback and assessment questionnaires from users across 23 time zones. High Wysa users showed significantly greater improvement in major depression symptoms with a moderate effect than low users, with 67.7% reporting the app as helpful and encouraging. The authors noted that these results aligned with prior RCTs on conversational agents for mental well-being [32,33] and emphasized that the naturalistic design offered scalable, real-time insights into engagement and effectiveness.

    A single-session “mini-course” version of the Tessa chatbot, focused on eating disorders, was studied in users prompted via social media searches [34]. This mini-course improved body image (moderate effect) and motivation to change (small effect), but the study was not fully naturalistic due to researcher screening before mini-course access. When Tessa was deployed in a real-world US setting in June 2023, it shifted from rule-based responses to offering “off-script” dieting and weight-loss advice, which can be very harmful in eating disorder contexts. This issue was reported by users, not researchers, and attracted global attention [35,36]. Since 2023, Tessa has not been implemented in any public forums to the best of our knowledge. Overall, real-world evaluations of body image– or eating disorder–specific chatbots and mental health chatbots in general remain scarce.

    The objective of the current study was to conduct a novel real-world evaluation of our body image and eating disorder education chatbot, JEM, in Australia and Canada, the 2 countries (representing 2 continents) that have hosted a deployment of the chatbot to date. Exploring both settings provided an opportunity to examine whether similar patterns of user experiences and outcomes were observed across different implementation contexts.

    Given the exploratory nature of this real-world evaluation, no a priori directional hypotheses were specified. However, the microinterventions were grounded in CBT, ACT, and mindfulness approaches and were examined as within-session pre-post changes in self-reported mood and body image satisfaction following engagement with the microinterventions. The primary aim of the study was to examine user satisfaction and experiences and explore within-session pre-post changes in body image satisfaction and mood for the microinterventions. No specific eating disorder symptoms or behavioral outcomes were assessed. A secondary aim was to explore and describe any differences between the Australian and Canadian JEM deployments.


    Ethical Considerations

    This study was approved by the Monash University Human Research Ethics Committee (ID 26129), which encompassed research conducted in both Australia and Canada. Chatbot users were offered the opportunity to complete web-based surveys throughout their conversation with the JEM chatbot; however, the surveys were framed as completely optional, and users could ignore the survey and continue using the chatbot without consequence. When users clicked on a survey weblink, they were presented with detailed study information. Users were informed of their right to withdraw from the survey at any time without any negative impacts. After the detailed information, users were asked to consent by clicking on a button that they had read the information and consented to proceeding to the survey. For users aged 13 to 15 years, ethics-approved procedures required them to confirm that they had obtained parental or guardian consent via an online acknowledgment. Given the fully anonymous and low-risk nature of the research, this approach was deemed appropriate and proportionate by the approving committee. The lower age threshold of 13 years was selected in line with common minimum ages for digital platform access and social media use and to capture early adolescent users who may engage with online eating disorder–related content. Users aged 16 and 17, like those aged 18 and over, were considered to be sufficiently mature to consent for themselves without parent or guardian consent required according to our ethics approval. Users completed the surveys anonymously—they were not required to provide their name or any potentially identifying information at any point. No compensation was offered for completing surveys or using the chatbot.

    Study Design

    This study used a naturalistic mixed methods design comprising 2 complementary data structures: (1) cross-sectional user evaluation surveys and (2) within-session pre-post microintervention assessments embedded within chatbot interactions. A minimum of 34 completed responses per microintervention was used as a pragmatic threshold intended to support estimation of moderate effect sizes [37], consistent with prior work [31,34], and the exploratory nature of this real-world design [38].

    Participants

    Participants were individuals aged 13 years and older residing in Australia or Canada who engaged with the JEM chatbot from January 2023 to June 2023 in Australia or September 2024 to March 2025 in Canada. Geographic location was based on self-reported residence and deployment platform context within an anonymous web-based design. These time periods covered the first 6 months of JEM deployment in each country for optimal comparability, particularly in terms of the novelty of the chatbot service to the users in the respective countries [39]. In this 6-month period, system-level usage included 6862 total chatbot sessions and 39,184 interactions in the Australian deployment and 1332 sessions and 5370 interactions for the Canadian deployment according to Google Dialogflow analytics definitions [40], which are distinct from individual-level survey respondents analyzed in this study. Study participants were recruited from individuals already engaging with the JEM chatbot. All surveys (overall evaluation and microintervention surveys) were open surveys, and a self-selection sampling method was used, representing a convenience sample. No direct contact with potential participants was made by any members of the research team. Recruitment occurred in writing through the chatbot’s “Provide Feedback/Feedback Survey” and “Skill Survey” conversation options, which users could select voluntarily. These options led to web-based surveys hosted by Qualtrics [41]. Surveys were administered with standard Qualtrics platform-level fraud-prevention settings applied to reduce duplicate responses (eg, cookie-based controls) [42,43], although these do not fully eliminate the possibility of repeat submissions in anonymous online settings. Only completed questionnaires were analyzed. Due to the fully anonymous design of the chatbot and surveys, individual users could not be tracked across sessions or timepoints. Therefore, engagement is reported at the session level, and survey participation represents voluntary, self-selected responses from the overall user base (see Multimedia Appendix 1 for aggregate study flow). For users who initiated the overall survey, completion rates were 80.1%, and for users who initiated microintervention surveys, completion rates ranged from 70.2% to 75.0%. There were no significant differences between the Australian and Canadian chatbot deployments for these completion rates (all Ps>.05).

    Intervention

    JEM was a web-based and rule-based chatbot powered by the Google Dialogflow platform [23]. The chatbot was available 24 hours per day, 7 days a week to users in Australia via a Monash University website (Figure 1) and in Canada via the National Eating Disorder Information Centre website (Figure 2). These deployment characteristics are reported to contextualize accessibility and real-world use; however, the primary focus of the intervention is the structure and content of the psychoeducational and microintervention pathways delivered within the chatbot. The chatbot was free of charge to access and users could converse with the chatbot as long as they wished to do so. There was no login or authentication required to use JEM, and usage was completely anonymous. Users were cautioned on these websites that the JEM chatbot was not monitored by any humans, and if they were experiencing an emergency, they should contact country-specific emergency services with contact details provided.

    ‎
    Figure 1. Australian version of the JEM chatbot hosted on a Monash University webpage. Users clicked on the blue speech bubble icon to start the conversation.
    ‎
    Figure 2. Canadian version of the JEM chatbot hosted on the National Eating Disorder Information Centre website. Users clicked on the purple JEM icon to start the conversation.

    The chatbot’s conversation content was based on evidence-based information or interventions for eating disorders, specifically psychoeducation, CBT, ACT, and mindfulness (for detailed information, see Beilharz et al [21]). Owing to the short conversational style of the JEM chatbot, the microinterventions taught by JEM were carefully selected, such as for CBT, education on cognitive distortions, for ACT, cognitive defusion exercises, and mindful breathing for mindfulness (see Figures 3 and 4 as examples). All microinterventions were estimated to take less than 5 minutes to complete. The chatbot operated through a rule-based decision-tree structure, where user inputs were matched to predefined intents that triggered specific conversational pathways. Users engaged with JEM via either free-text inputs or menu-based options, which routed them to psychoeducation content or presented a set of available microintervention scripts. The 10 microintervention scripts were discrete, standardized conversational pathways that users could actively select from a menu of available skills. Each microintervention followed a fixed sequence of prompts and responses with no within-script branching. These scripts were delivered in multimodal formats, including text, images, audio, and video. Users self-selected into microinterventions during their interaction with the chatbot, and each script could be completed within a single session or revisited multiple times. Completion was operationalized as reaching the end of the scripted conversational pathway.

    ‎
    Figure 3. Example conversation from the JEM chatbot for the “Managing Emotions” microintervention (Canadian version).

    There were no major conceptual differences in intervention content and technical delivery between the Australian and Canadian deployments. In terms of safety, natural language processing was used for JEM to detect user-typed risk phrases such as “I want to kill myself,” and “I wish I was dead,” and JEM responded with 24/7 crisis support services specific to Australia or Canada. Specifically, risk detection was implemented using Google Dialogflow’s intent classification and entity extraction [44] to identify free-text inputs indicative of self-harm and suicidality. When a high-risk intent or entity was detected, the system immediately triggered a predefined response flow that bypassed standard conversational pathways and directed users to country-specific 24/7 crisis support services. There was also a “Get Urgent Help Now” option for users to select where the chatbot responded with these same crisis support services. If a user was to type a prompt that the chatbot could not match to a preprogrammed answer, the chatbot responded with “I’m just a simple bot so I’m still learning how to respond to your typed message. Try using my buttons to have a conversation with me” along with crisis support service contacts. There were no adverse events reported by users during the study period.

    ‎
    Figure 4. Example conversation from the JEM chatbot for the “Feeling Worthy” microintervention (Australian version).

    Measures

    Surveys

    There were 2 types of web-based surveys, hosted by Qualtrics [41], where the survey weblinks were included within the JEM chatbot’s conversation: (1) overall evaluation survey accessed via the “Provide Feedback” option and (2) skill surveys presented for every microintervention conversation (10 in total). There were no “back” options on any of the surveys for people to change their earlier responses.

    Overall Evaluation Survey

    The content and structure of this survey were based on our previous chatbot co-design research [20,21]. These items were designed as brief, study-specific indicators of user experience within an embedded chatbot context and were not intended as stand-alone psychometric instruments. In the first section of the survey, demographic characteristics were asked, particularly age, gender identity, LGBTIQA+ community membership (lesbian, gay, bisexual, transgender, intersex, queer, asexual people, or people otherwise diverse in gender, sexual orientation, and/or innate variations of sex characteristics), and ethnicity. Note that ethnicity was asked in a different format in Australia compared to Canada to suit the standard ethnicity demographics of the country [45]. The second section of the survey included categorical items addressing (1) reasons for using the chatbot, (2) whether the user found the information they were looking for, and (3) whether they intended to seek further support. The third section involved a single-item user satisfaction sliding scale ranging from 0 (“not at all satisfied”) to 100 (“completely satisfied”). This satisfaction item was followed by an optional open-text question: “Please let us know your reason(s) for your satisfaction rating.” The fourth and final section of the survey used the validated System Usability Scale (SUS) [46], which comprised 10 items that reflected various statements related to system usability (eg, “I thought the system was easy to use”). Responses were recorded on a 5-point Likert scale ranging from 1 (“strongly disagree”) to 5 (“strongly agree”). The SUS score yielded a single number between 0 and 100, with higher scores representing higher perceived usability of the system. The SUS demonstrated high reliability in this study’s sample (Cronbach α=0.89 for both Australian and Canadian samples).

    Microintervention Survey

    For every microintervention included in the JEM chatbot (10 in total, see Beilharz et al [21] for comprehensive descriptions), there was a web-based survey offered. If the user decided to complete the survey, they were asked for their current mood ratings for how happy, confident, anxious, and depressed they felt, with each being rated on a visual analog scale ranging from 0 (“not at all”) to 100 (“very much”). The respondents were then asked to similarly rate for body image satisfaction on visual analog scales ranging from 0 (“not at all”) to 100 (“very much”), specifically physical attractiveness, body size satisfaction, and body shape satisfaction. Note that we did not ask demographic characteristics of respondents in the microintervention surveys. Single-item visual analog scales were used to assess momentary mood and body satisfaction to minimize respondent burden and enable repeated, brief assessments within a chatbot-based design. Single-item visual analog scale measures are widely used for capturing transient, state-level affective and body image experiences and have been shown to be sensitive to within-person changes over time [47-50]. Importantly, prior research indicates that single-item measures of body satisfaction are responsive to short-term changes in both observational and online intervention contexts [51-53]. Respondents were instructed to complete the microintervention and then return to the survey immediately afterward, where they were then presented with the same mood and body image satisfaction visual analog scales again. Thus, the before and after microintervention ratings were completed within the same survey for each microintervention.

    Data Analysis

    The Statistical Package for Social Sciences (SPSS; version 31; IBM Corp) [54] was used for quantitative statistical analysis. Descriptive statistics were used to analyze demographic characteristics and user satisfaction and experiences. To explore if there were differences between countries, 2-tailed independent samples t tests were used for continuous data and Fisher’s exact tests for categorical data (2-tailed). To examine within-session pre-post microintervention changes in mood and body image, Cohen d was calculated as a measure of effect size [37], with 95% CIs reported. Analyses were conducted at the survey or session level due to the fully anonymous design of the chatbot, which did not allow the linkage of responses across users or timepoints. Any missing data were handled with listwise deletion, acknowledging that missingness in optional, anonymous web-based surveys may not be completely at random. Given the exploratory and hypothesis-generating nature of these analyses, no formal correction for multiple testing was applied [55]. Applying such corrections in this context may have increased the risk of type 2 error and obscured potentially meaningful signals warranting further investigation. The results were therefore interpreted cautiously, acknowledging the increased risk of type 1 error associated with multiple comparisons. The open-ended qualitative text responses addressing reasons for chatbot satisfaction ratings were analyzed using an abbreviated content analysis approach [56], given the brevity of the responses (30 words maximum). GS conducted initial coding, with HH independently reviewing and refining codes. All authors agreed on the final content interpretations.


    User Characteristics

    Across all analyses, reported Ns refer to completed survey instances (survey responses). As shown in Table 1, from the overall evaluation surveys, respondents were, on average, in their early 20s, but age distribution varied substantially, particularly in the Australian sample. The respondents predominantly identified as girls or women, followed by nonbinary individuals. Around two-thirds of respondents did not identify as LGBTIQA+; however, when they did, queer was the most common identity. There were no differences between Australian and Canadian-based respondents for any of these demographic characteristics (all P values >.05). The majority of respondents were European or White in Canada, and around a quarter identified as culturally and linguistically diverse (CALD) in Australia. Owing to the different style of questioning required, ethnicity was not compared between countries.

    Table 1. Respondent demographic characteristics for overall evaluation survey by country (N=230 total).
    Demographic characteristicAustralia (n=122)Canada (n=108)
    Age (y), mean (SD; range)21.0 (12.0; 13‐64)22.8 (6.8; 14‐43)
    Gender, n (%)
     Girl or woman99 (81.1)87 (80.6)
     Boy or man7 (5.7)6 (5.6)
     Nonbinary14 (11.5)12 (11.1)
     Transgender woman0 (0)0 (0)
     Transgender man1 (0.8)0 (0)
     Prefer not to answer1 (0.8)3 (2.8)
    Ethnicity
    Australia CALDa, n (%)
      No89 (73.0)—b
      Yes31 (25.4)—
      Prefer not to answer2 (1.6)—
    Canada, n (%)
      European or White—81 (75.0)
      African or Black—3 (2.8)
      Latin, South or Central American—3 (2.8)
      East Asian—3 (2.8)
      “Mixed”—6 (5.6)
      Prefer not to answer—12 (11.1)
    LGBTIQA+c n (%)
     No81 (66.4)72 (66.7)
     Bisexual4 (3.3)6 (5.6)
     Lesbian5 (4.1)3 (2.8)
     Queer17 (13.9)15 (13.9)
     Gay10 (8.2)6 (5.6)
     Asexual0 (0.0)3 (2.8)
     Prefer not to answer5 (4.1)3 (2.8)

    aCALD: culturally and linguistically diverse.

    bEthnicity was assessed using country-specific response categories. Blank cells indicate categories that were not applicable to that country.

    cLGBTIQA+: lesbian, gay, bisexual, transgender, intersex, queer, asexual people, or people otherwise diverse in gender, sexual orientation, and/or innate variations of sex characteristics.

    Experiences and Satisfaction

    As indicated in Table 2 from the overall evaluation surveys, the most common reasons in both countries for using JEM chatbot were that the respondents thought they may have an eating disorder, had issues finding information, and were simply curious about the chatbot. Respondents found the information they were looking for the vast majority of the time, and the majority intended to seek or maybe seek further support after using the chatbot. The overall satisfaction rating for the JEM chatbot was high on average, and the SUS rating was in the “excellent” category for both Australia and Canada [46]. There were no country-based significant differences for any of these findings (all P values >.05).

    For the open-ended qualitative responses to reasons for the satisfaction ratings, this question was completed by a self-selected subset of respondents (Australia n=16; Canada n=12), which did not significantly differ between countries (all P values>.05). Given the lower response rates, responses were analyzed in aggregate rather than by country (Table 3). The response themes were mostly positive in nature, with the most common focused on JEM chatbot providing helpful information and being easy to use.

    Table 2. Responses to overall evaluation survey by country (N=230 total).
    MeasureAustralia (n=122)Canada (n=108)
    Reasons for using the chatbot, n (%)a
     May have an eating disorder71 (58.2)61 (56.5)
     Trouble finding information via other sources32 (26.2)29 (26.9)
     Curious about the chatbot29 (23.8)27 (25.0)
     Chatbots do not judge13 (10.7)12 (11.1)
     Not ready to speak to a real person10 (8.2)9 (8.3)
    Find information, n (%)
     Yes112 (91.8)100 (92.6)
     No10 (8.2)8 (7.4)
    Seeking further support, n (%)
     Yes79 (64.8)77 (71.3)
     Maybe15 (12.3)14 (13.0)
     No14 (11.5)12 (11.1)
     I already have support14 (11.5)5 (4.6)
    Satisfaction rating, mean (SD; range)76.1 (22.7; 20.0‐100)78.8 (14.3; 30.0‐100)
    SUSb rating, mean (SD; range)86.5 (16.9; 20.0‐100)89.5 (11.6; 40.0‐100)

    aPercentages add up to greater than 100% as respondents could choose multiple responses.

    bSUS: System Usability Scale.

    Table 3. Themes for open-ended qualitative responses for reasons for satisfaction ratings by country (N=28 total).
    ThemeCategoryTotal, n (%)aAustralia, nCanada, n
    Chatbot provided helpful informationPositive18 (64.3)108
    Chatbot was easy to usePositive16 (57.1)106
    JEM character was fun or cutePositive7 (25.0)34
    Prefer a chatbot over a helplinePositive6 (21.4)60
    Chatbot had a friendly tonePositive4 (14.3)04
    Difficult to find the information soughtNegative2 (7.1)20
    Chatbot conversation was targeted at too young an audienceNegative1 (3.6)01

    aPercentages add up to greater than 100% as qualitative responses could be classified into multiple themes.

    Changes to Mood and Body Image

    Four of the 10 total microinterventions included in the JEM chatbot’s conversation yielded sufficient completed survey responses (n≥34) from either country to conduct suitable within-session pre- or postchange analyses (Table 4). These 4 microinterventions were (1) Managing unhelpful thoughts (CBT and ACT strategies), (2) Managing emotions (ACT strategies), (3) Challenging beauty (CBT strategies), and (4) Feeling worthy (CBT strategies). Sample sizes for individual microintervention analyses varied and were modest for some subgroups; therefore, the findings should be interpreted with appropriate caution.

    Across microinterventions, the within-session pre-post ratings were broadly similar in Australia and Canada, with effect sizes ranging from very small to large. The Managing unhelpful thoughts microintervention showed very small or small effect sizes across most measures, with moderate effects observed for anxiety. The Managing emotions microintervention showed the largest within-session effect sizes for depression, with moderate effects observed for body size satisfaction and the remaining measures showing very small-to-small effects. The Challenging beauty microintervention showed very small-to-moderate effects across measures, with confidence and physical attractiveness showing the largest effect sizes. The Feeling worthy microintervention showed very small-to-moderate effects across measures, with confidence, physical attractiveness, body size satisfaction, and body shape satisfaction showing the largest effect sizes.

    Table 4. Responses to mood and body image measures before and after microintervention surveys by country.
    MeasureAustraliaCanada
    Before, mean (SD)After, mean (SD)Cohen d (95% CI)Before, mean (SD)After, mean (SD)Cohen d (95% CI)
    Microintervention: Managing unhelpful thoughts (n=146 for Australia and n=130 for Canada)
    Happy28.9 (21.9)29.4 (24.2)0.03 (−0.14 to 0.19)29.7 (21.7)29.2 (23.1)−0.03 (−0.20 to 0.14)
    Confident20.0 (20.0)23.4 (21.8)0.20 (0.04 to 0.37)20.5 (19.8)23.0 (20.7)0.16 (−0.01 to 0.34)
    Anxious63.0 (28.7)53.9 (30.9)−0.45 (−0.62 to −0.23)64.1 (28.3)54.4 (30.4)−0.53 (−0.70 to −0.34)
    Depressed50.4 (31.4)45.3 (32.0)−0.25 (−0.42 to −0.09)51.2 (30.7)45.7 (31.7)−0.29 (−0.46 to −0.11)
    Physically attractive21.9 (24.5)22.9 (24.3)0.06 (−0.10 to 0.23)22.1 (24.4)22.2 (23.0)0.00 (−0.17 to 0.17)
    Satisfied body size17.8 (24.0)20.4 (24.6)0.20 (0.03 to 0.37)18.3 (24.0)20.0 (23.2)0.15 (−0.02 to 0.33)
    Satisfied body shape17.8 (24.5)20.2 (26.1)0.18 (0.01 to 0.34)18.2 (24.6)19.7 (25.2)0.14 (−0.04 to 0.31)
    Microintervention: Managing emotions (n=34 for Australia and n=44 for Canada)
    Happy36.6 (21.8)38.3 (24.8)0.14 (−0.20 to 0.48)35.5 (21.2)38.9 (23.6)0.22 (−0.09 to 0.52)
    Confident24.2 (14.8)26.0 (19.5)0.12 (−0.22 to 0.46)23.6 (15.6)27.3 (20.2)0.22 (−0.08 to 0.53)
    Anxious65.6 (26.4)56.2 (28.0)−0.38 (−0.73 to −0.03)64.5 (26.9)53.7 (28.3)−0.43 (−0.73 to −0.11)
    Depressed48.8 (29.9)34.1 (24.5)−0.78 (−1.13 to −0.38)48.4 (29.1)31.9 (22.8)−0.79 (−1.12 to −0.44)
    Physically attractive22.2 (22.8)23.5 (26.3)0.12 (−0.22 to 0.46)20.9 (22.5)22.8 (25.7)0.18 (−0.12 to 0.48)
    Satisfied body size13.1 (17.6)17.7 (21.7)0.43 (0.07 to 0.78)13.1 (17.1)18.1 (20.8)0.51 (0.19 to 0.82)
    Satisfied body shape17.5 (21.9)19.2 (21.8)0.19 (−0.15 to 0.54)18.5 (22.5)20.1 (22.9)0.19 (−0.11 to 0.49)
    Microintervention: Challenging beauty (n=50 for Australia and n=40 for Canada)
    Happy40.5 (30.5)43.6 (31.4)0.15 (−0.13 to 0.43)38.8 (29.2)41.5 (29.5)0.13 (−0.19 to 0.44)
    Confident23.5 (25.9)33.6 (29.6)0.47 (0.18 to 0.76)25.2 (26.4)32.2 (29.2)0.38 (0.06 to 0.70)
    Anxious59.4 (31.1)47.7 (33.5)−0.44 (−0.73 to −0.15)60.5 (30.5)52.9 (33.0)−0.32 (−0.63 to 0.00)
    Depressed46.7 (33.5)40.4 (29.1)−0.31 (−0.59 to −0.03)49.2 (30.6)43.7 (29.4)−0.25 (−0.56 to 0.07)
    Physically attractive20.4 (25.5)29.5 (32.1)0.44 (0.15 to 0.73)21.0 (26.6)28.6 (30.9)0.42 (0.09 to 0.74)
    Satisfied body size20.6 (27.7)27.6 (32.7)0.33 (0.04 to 0.62)21.2 (29.8)26.3 (31.9)0.26 (−0.06 to 0.58)
    Satisfied body shape23.0 (27.7)28.3 (30.9)0.24 (−0.05 to 0.52)22.6 (28.7)27.1 (29.1)0.21 (−0.11 to 0.53)
    Microintervention: Feeling worthy (n=36 for Australia and n=39 for Canada)
    Happy28.4 (21.1)31.9 (24.0)0.15 (−0.18 to 0.47)24.3 (20.0)31.0 (25.6)0.23 (−0.09 to 0.54)
    Confident17.4 (16.1)28.2 (22.5)0.46 (0.11 to 0.80)14.5 (15.5)28.3 (25.1)0.50 (0.16 to 0.83)
    Anxious63.5 (27.4)51.2 (33.0)−0.37 (−0.71 to −0.03)65.4 (29.3)55.1 (34.5)−0.27 (−0.60 to 0.05)
    Depressed61.1 (26.1)58.4 (29.0)−0.10 (−0.43 to 0.23)59.5 (26.4)58.4 (29.6)−0.03 (−0.35 to 0.03)
    Physically attractive12.6 (16.8)21.7 (26.0)0.47 (0.11 to 0.81)10.0 (11.8)21.7 (26.9)0.47 (0.13 to 0.80)
    Satisfied body size8.9 (12.3)16.8 (21.4)0.40 (0.05 to 0.72)7.8 (11.5)19.2 (24.7)0.45 (0.11 to 0.78)
    Satisfied body shape10.2 (13.5)17.4 (22.8)0.41 (0.06 to 0.74)11.2 (15.5)20.7 (27.3)0.42 (0.09 to 0.74)

    Principal Findings

    This study provides one of the first real-world evaluations of a mental health–focused chatbot deployed at scale across 2 countries and continents, providing insight into chatbot use under naturalistic conditions. Overall, the findings indicated that the JEM chatbot was well accepted by respondents who completed surveys in both Australia and Canada, demonstrated excellent usability, and showed immediate within-session differences in mood and body image ratings. Satisfaction was high, information needs were largely met, and most survey respondents reported an intention to seek or consider further support after using the chatbot, suggesting that JEM may function as a supportive resource that could potentially facilitate consideration of further support.

    Across microinterventions, patterns of within-session pre-post ratings were broadly similar in Australia and Canada, with effect sizes generally ranging from very small to large. While the magnitude of effect sizes varied across microinterventions and outcomes, the largest effects were typically observed for depression, anxiety, confidence, and body size satisfaction outcomes. These findings reflect immediate within-session pre-post changes and should not be interpreted as evidence of causal or sustained intervention effects.

    In contrast, effect sizes for happiness ratings were generally very small across the microinterventions included in the analyses. This may reflect the brief nature of the interventions, which are designed to reduce distress and maladaptive cognitions rather than induce immediate positive affect, and is consistent with theoretical expectations of CBT and ACT informed strategies [57-59].

    Gender identity beyond the binary was represented in the sample, with approximately 11% of respondents identifying as nonbinary. All study analyses included these respondents within the overall sample rather than stratifying by gender. Overall, JEM’s nongendered and inclusive framing of content [21] suggests potential applicability across gender identities, although future work should directly examine differential experiences among gender diverse users, particularly their responses to psychotherapeutic microinterventions.

    Comparison With Prior Work

    These findings extend the very limited literature on eating disorder– and body image–specific intervention chatbots, particularly in real-world contexts. Previous RCTs have demonstrated efficacy for Tessa, Topity, and ED ESSI under controlled conditions [15,16,19], while other chatbots have been evaluated only partially or at early stages of development [17,18]. This study extends this work by demonstrating that a rule-based chatbot can be deployed with high user satisfaction and usability in naturalistic settings. Furthermore, the patterns of within-session pre-post differences observed among JEM users aligned in direction with findings reported in more controlled trials, while recognizing the differences in study design.

    Compared with prior naturalistic evaluations of mental health chatbots such as Wysa [31], this study focused on a more specific population and set of outcomes. The patterns of within-session pre-post differences in anxiety, depression, and body image satisfaction observed among JEM survey respondents were mostly consistent with those reported in broader mental health chatbot evaluations. These observations are consistent with the literature suggesting that well-designed conversational agents may be associated with positive user-reported outcomes in real-world observational settings [31,60] and align with broader discussions about the promise and fragility of AI-mediated mental health care in real-world settings [61].

    Importantly, the present findings contrast with concerns raised following the real-world deployment of the Tessa chatbot in 2023 in the United States, where the chatbot deviated from its rule-based design and provided dieting and weight loss advice [35]. No adverse events or harmful outputs were reported by users during the study period. JEM’s rule-based architecture and co-designed content were intended to minimize such risks; however, all safety outcomes were not assessed in this study.

    The very small effects on happiness from microinterventions also align with previous digital intervention research, where reductions in negative symptoms are often more readily observed than increases in positive affect, particularly following brief, single-session interventions [59,62]. Together, these findings suggest that JEM performs in a manner consistent with existing evidence while addressing a notable gap in real-world evaluations of body image– and eating disorder–specific chatbots.

    Recent RCTs of generative AI-based conversational agents further suggest both the promise and the complexity of chatbot-delivered mental health interventions, reinforcing the importance of safety-focused design and careful real-world evaluation [63]. While the vast majority of qualitative feedback in this study was positive, with respondents reporting helpfulness, ease of use, and an engaging conversational style, a minority had negative feedback. More broadly, rule-based chatbot systems are often perceived as less flexible than emerging large language model (LLM)–based approaches, which offer more naturalistic dialogue [64]. This highlights an important design trade-off between maintaining clinically safe, structured responses and meeting evolving user expectations for more adaptive and personalized conversational experiences. Future JEM iterations could explore hybrid architectures that combine rule-based safety constraints with controlled generative or retrieval-based components to enhance conversational flexibility while maintaining clinical safety and content fidelity [65].

    Limitations and Future Research

    Several limitations should be acknowledged for this study. First, the absence of a control group precludes causal inference and means changes cannot be definitively attributed to the chatbot alone. Second, outcomes were assessed immediately before and after individual microinterventions, limiting conclusions about the durability of effects. The immediate pre- and postintervention design may also be susceptible to demand characteristics and expectancy effects. However, differences were not uniform across outcomes or microinterventions, suggesting variability in within-session responses across measures and intervention content. Third, sample sizes varied across microinterventions, with 4 yielding sufficient data for analysis, which may have influenced the pattern of the findings based on differential engagement. Additionally, reliance on self-report measures and voluntary survey completion introduces potential response bias and means that the results reflect only a subset of users. Due to the fully anonymous design of JEM, which did not include login, authentication, or persistent user identifiers, individuals could not be tracked across sessions or over time. As a result, it was not possible to construct a participant-level CONSORT-EHEALTH (Consolidated Standards of Reporting Trials of Electronic and Mobile Health Applications and Online Telehealth) style flow diagram [66] tracking individuals through the chatbot content or determine precisely where attrition occurred within the chatbot journey. Engagement data are therefore reported via text and diagram at the session level, limiting the assessment of repeat usage and longitudinal engagement. Furthermore, this study did not assess changes in specific eating disorder symptoms, eating behaviors, or clinical outcomes.

    Future research should examine longer-term outcomes associated with repeated or sustained chatbot use, including whether these immediate within-session differences are associated with longer-term outcomes, such as eating disorder risk or engagement with formal care. Hybrid designs combining real-world deployment with comparison conditions may help balance ecological validity with internal rigor. Further investigation into engagement patterns, personalization, and differential effects across demographic subgroups is also warranted.

    Conclusions

    This study demonstrated that a co-designed evidence-based chatbot, JEM, can be deployed at scale across multiple countries, achieving high usability and user satisfaction. Immediate within-session differences in mood and body image ratings were observed following the completion of the microinterventions investigated. However, further controlled research is needed, and the longer-term effects are yet to be determined. The study findings were broadly similar across Australia and Canada. These results extend the limited real-world evidence for mental health and body image– or eating disorder–specific chatbots, demonstrating that rule-based systems with co-designed content can be acceptable to users. As digital mental health interventions evolve, future research should explore how different approaches, including rule-based and generative AI chatbots, can complement each other to provide scalable, safe, and personalized support. Collectively, these findings suggest that chatbots such as JEM may represent a feasible and acceptable low-barrier resource for individuals seeking information and support related to body image and eating disorder concerns.

    Acknowledgments

    The authors would like to thank the JEM chatbot users in Australia and Canada for their valuable study involvement.

    ChatGPT (OpenAI) was used to assist with language editing and improving clarity and flow of the manuscript. The authors take full responsibility for the content.

    Funding

    The authors declared no financial support was received for this work.

    Data Availability

    The datasets generated and analyzed during the current study are not publicly available under the participant confidentiality conditions of ethics approval from the Monash University Human Research Ethics Committee (ID 26129). The corresponding author can be contacted with queries about the data.

    Authors' Contributions

    Conceptualization: GS

    Data curation: GS, HH

    Formal analysis: GS, HH

    Investigation: GS

    Methodology: GS, SM, ET

    Project administration: GS, SM, ET

    Software: HH

    Supervision: GS

    Writing – original draft: GS

    Writing – review and editing: GS, SM, ET, HH

    Conflicts of Interest

    GS is a Section Editor for JMIR Mental Health. The JEM chatbot was originally developed at Monash University and is now owned and managed by GS, including the JEM® registered trademark. All other authors declare no conflicts of interest.

    Multimedia Appendix 1

    Aggregate study flow for the Australian (A) and Canadian (B) JEM chatbot deployments. Ten microinterventions were available within JEM, of which 4 generated sufficient completed surveys to meet the prespecified threshold for analysis.

    PNG File, 44 KB
    1. Diagnostic and Statistical Manual of Mental Disorders. 5th ed. American Psychiatric Association Publishing; 2022. ISBN: 978-0-89042-576-3
    2. Hay P, Aouad P, Le A, et al. Epidemiology of eating disorders: population, prevalence, disease burden and quality of life informing public policy in Australia-a rapid review. J Eat Disord. Feb 15, 2023;11(1):23. [CrossRef] [Medline]
    3. Qian J, Wu Y, Liu F, et al. An update on the prevalence of eating disorders in the general population: a systematic review and meta-analysis. Eat Weight Disord. Mar 2022;27(2):415-428. [CrossRef] [Medline]
    4. Shagar PS, Harris N, Boddy J, Donovan CL. The relationship between body image concerns and weight-related behaviours of adolescents and emerging adults: a systematic review. Behav change. Dec 2017;34(4):208-252. [CrossRef]
    5. Deloitte Access Economics. Paying the price report. Butterfly Foundation; 2024. URL: https:/​/butterfly.​org.au/​wp-content/​uploads/​2024/​10/​deloitte-au-eco-paying-the-price-second-edition-180724-new-Oct-24.​pdf? [Accessed 2026-07-03]
    6. Obeid N, Coelho JS, Booij L, et al. Estimating additional health and social costs in eating disorder care for young people during the COVID-19 pandemic: implications for surveillance and system transformation. J Eat Disord. Apr 26, 2024;12(1):52. [CrossRef] [Medline]
    7. Obeid N, Silva-Roy P, Booij L, Coelho JS, Dimitropoulos G, Katzman DK. The financial and social impacts of the COVID-19 pandemic on youth with eating disorders, their families, clinicians and the mental health system: a mixed methods cost analysis. J Eat Disord. Mar 29, 2024;12(1):43. [CrossRef] [Medline]
    8. Löchner J, Carlbring P, Schuller B, Torous J, Sander LB. Digital interventions in mental health: an overview and future perspectives. Internet Interv. Jun 2025;40:100824. [CrossRef] [Medline]
    9. Torous J, Linardon J, Goldberg SB, et al. The evolving field of digital mental health: current evidence and implementation issues for smartphone apps, generative artificial intelligence, and virtual reality. World Psychiatry. Jun 2025;24(2):156-174. [CrossRef] [Medline]
    10. Haque MDR, Rubya S. An overview of chatbot-based mobile mental health apps: insights from app description and user reviews. JMIR mHealth uHealth. May 22, 2023;11:e44838. [CrossRef] [Medline]
    11. He Y, Yang L, Qian C, et al. Conversational agent interventions for mental health problems: systematic review and meta-analysis of randomized controlled trials. J Med Internet Res. Apr 28, 2023;25(1):e43862. [CrossRef] [Medline]
    12. Vaidyam AN, Wisniewski H, Halamka JD, Kashavan MS, Torous JB. Chatbots and conversational agents in mental health: a review of the psychiatric landscape. Can J Psychiatry. Jul 2019;64(7):456-464. [CrossRef] [Medline]
    13. Feng X, Tian L, Ho GWK, Yorke J, Hui V. The effectiveness of AI chatbots in alleviating mental distress and promoting health behaviors among adolescents and young adults: systematic review and meta-analysis. J Med Internet Res. Nov 26, 2025;27:e79850. [CrossRef] [Medline]
    14. Fu L, Burns R, Xie Y, et al. The development and use of AI chatbots for health behavior change: scoping review. J Med Internet Res. Jan 28, 2026;28:e79677. [CrossRef] [Medline]
    15. Fitzsimmons-Craft EE, Chan WW, Smith AC, et al. Effectiveness of a chatbot for eating disorders prevention: a randomized clinical trial. Int J Eat Disord. Mar 2022;55(3):343-353. [CrossRef] [Medline]
    16. Matheson EL, Smith HG, Amaral ACS, et al. Using chatbot technology to improve Brazilian adolescents’ body image and mental health at scale: randomized controlled trial. JMIR mHealth uHealth. Jun 19, 2023;11:e39934. [CrossRef] [Medline]
    17. Shah J, DePietro B, D’Adamo L, et al. Development and usability testing of a chatbot to promote mental health services use among individuals with eating disorders following screening. Int J Eat Disord. Sep 2022;55(9):1229-1244. [CrossRef] [Medline]
    18. Gullo NA, Goldberg J, Howe CP, et al. User-centered development of a chatbot for diverse adolescents at high risk for eating disorders. Int J Eat Disord. Apr 2026;59(4):724-735. [CrossRef] [Medline]
    19. Sharp G, Dwyer B, Randhawa A, McGrath I, Hu H. The effectiveness of a chatbot single-session intervention for people on waitlists for eating disorder treatment: randomized controlled trial. J Med Internet Res. May 21, 2025;27:e70874. [CrossRef] [Medline]
    20. Sharp G, Dwyer B, Xie J, et al. Co-design of a single session intervention chatbot for people on waitlists for eating disorder treatment: a qualitative interview and workshop study. J Eat Disord. Mar 11, 2025;13(1):46. [CrossRef] [Medline]
    21. Beilharz F, Sukunesan S, Rossell SL, Kulkarni J, Sharp G. Development of a positive body image chatbot (KIT) with young people and parents/carers: qualitative focus group study. J Med Internet Res. Jun 16, 2021;23(6):e27807. [CrossRef] [Medline]
    22. Fitzsimmons-Craft EE, Rackoff GN, Shah J, et al. Effects of chatbot components to facilitate mental health services use in individuals with eating disorders following online screening: an optimization randomized controlled trial. Int J Eat Disord. Nov 2024;57(11):2204-2216. [CrossRef] [Medline]
    23. Customer experience agent studio. Google Cloud. URL: https://cloud.google.com/products/conversational-agents?hl=en [Accessed 2025-12-30]
    24. Hart L, Phillipou A. COVID has presented unique challenges for people with eating disorders. They’ll need support beyond the pandemic. The Conversation. 2020. URL: https:/​/theconversation.​com/​covid-has-presented-unique-challenges-for-people-with-eating-disorders-theyll-need-support-beyond-the-pandemic-148903 [Accessed 2025-03-03]
    25. National mental health and wellbeing pandemic response plan. Australian Government, National Mental Health Commission; 2020. URL: https:/​/www.​mentalhealthcommission.gov.au/​sites/​default/​files/​2024-03/​national-mental-health-and-wellbeing-pandemic-response-plan.​pdf [Accessed 2026-07-03]
    26. Mohr DC, Schueller SM, Riley WT, et al. Trials of intervention principles: evaluation methods for evolving behavioral intervention technologies. J Med Internet Res. Jul 8, 2015;17(7):e166. [CrossRef] [Medline]
    27. Manole A, Cârciumaru R, Brînzaș R, Manole F. An exploratory investigation of chatbot applications in anxiety management: a focus on personalized interventions. Information. 2025;16(1):11. [CrossRef]
    28. Fan X, Chao D, Zhang Z, Wang D, Li X, Tian F. Utilization of self-diagnosis health chatbots in real-world settings: case study. J Med Internet Res. Jan 6, 2021;23(1):e19928. [CrossRef] [Medline]
    29. Nadarzynski T, Miles O, Cowie A, Ridge D. Acceptability of artificial intelligence (AI)-led chatbot services in healthcare: a mixed-methods study. Digit Health. 2019;5:2055207619871808. [CrossRef] [Medline]
    30. Jabir AI, Martinengo L, Lin X, Torous J, Subramaniam M, Tudor Car L. Evaluating conversational agents for mental health: scoping review of outcomes and outcome measurement instruments. J Med Internet Res. Apr 19, 2023;25(9):e44548. [CrossRef] [Medline]
    31. Inkster B, Sarda S, Subramanian V. An empathy-driven, conversational artificial intelligence agent (Wysa) for digital mental well-being: real-world data evaluation mixed-methods study. JMIR mHealth uHealth. Nov 23, 2018;6(11):e12106. [CrossRef] [Medline]
    32. Fitzpatrick KK, Darcy A, Vierhile M. Delivering cognitive behavior therapy to young adults with symptoms of depression and anxiety using a fully automated conversational agent (Woebot): a randomized controlled trial. JMIR Ment Health. Jun 6, 2017;4(2):e19. [CrossRef] [Medline]
    33. Ly KH, Ly AM, Andersson G. A fully automated conversational agent for promoting mental well-being: a pilot RCT using mixed methods. Internet Interv. Dec 2017;10(C):39-46. [CrossRef] [Medline]
    34. Nemesure MD, Park C, Morris RR, et al. Evaluating change in body image concerns following a single session digital intervention. Body Image. Mar 2023;44:64-68. [CrossRef] [Medline]
    35. Jargon J. A chatbot was designed to help prevent eating disorders. Then it gave dieting tips. The Wall Street Journal. 2023. URL: https://www.wsj.com/articles/eating-disorder-chatbot-ai-2aecb179#comments_sector [Accessed 2024-12-30]
    36. Sharp G, Torous J, West ML. Ethical challenges in AI approaches to eating disorders. J Med Internet Res. Aug 14, 2023;25:e50696. [CrossRef] [Medline]
    37. Cohen J. A power primer. Psychol Bull. Jul 1992;112(1):155-159. [CrossRef] [Medline]
    38. Faul F, Erdfelder E, Lang AG, Buchner A. G*Power 3: a flexible statistical power analysis program for the social, behavioral, and biomedical sciences. Behav Res Methods. May 2007;39(2):175-191. [CrossRef] [Medline]
    39. Talukdar N, Yu S. Breaking the psychological distance: the effect of immersive virtual reality on perceived novelty and user satisfaction. J Strateg Mark. Nov 16, 2024;32(8):1147-1171. [CrossRef]
    40. Analytics. Google Cloud. URL: https://docs.cloud.google.com/dialogflow/es/docs/analytics [Accessed 2025-12-30]
    41. Qualtrics. URL: https://www.qualtrics.com [Accessed 2025-12-30]
    42. Fraud detection. Qualtrics. 2026. URL: https://www.qualtrics.com/support/survey-platform/survey-module/survey-checker/fraud-detection [Accessed 2026-01-01]
    43. Prevent multiple submissions. Qualtrics. 2026. URL: https:/​/www.​qualtrics.com/​support/​survey-platform/​survey-module/​survey-options/​survey-protection/​#:~:text=Prevent%20Multiple%20Submissions,-Qtip%3A%20This%20option&text=This%20setting%20works%20by%20placing,them%20to%20take%20the%20survey [Accessed 2026-01-01]
    44. Intent matching. Google Cloud. URL: https:/​/docs.​cloud.google.com/​dialogflow/​es/​docs/​intents-matching#:~:text=When%20an%20end%2Duser%20writes,also%20known%20as%20intent%20classification [Accessed 2026-04-30]
    45. Guidance on demographic questions. McMaster University—Research & Innovation. URL: https:/​/research.​mcmaster.ca/​home/​support-for-researchers/​ethics/​mcmaster-research-ethics-board-mreb/​guidance-on-demographic-questions [Accessed 2026-01-01]
    46. Brooke J. SUS: A 'Quick and Dirty' Usability Scale. In: Jordan PW, Thomas B, Weerdmeester WA, McClelland AL, editors. Usability Evaluation in Industry. Taylor and Francis; 1996:189-194. ISBN: 9780429157011
    47. Birkeland R, Thompson JK, Herbozo S, Roehrig M, Cafri G, van den Berg P. Media exposure, mood, and body image dissatisfaction: an experimental test of person versus product priming. Body Image. Mar 2005;2(1):53-61. [CrossRef] [Medline]
    48. Tiggemann M, Zaccardo M. Mood and body dissatisfaction visual analogue scales [database record]. APA Psyc Tests. 2015. URL: https://psycnet.apa.org/doiLanding?doi=10.1037%2Ft47509-000 [Accessed 2026-07-16]
    49. Cash TF. Cognitive-behavioral perspectives on body image. In: Cash TF, Pruzinsky T, editors. Body Image: A Handbook of Theory, Research, and Clinical Practice. Guilford Press; 2002:269-276. ISBN: 9781572307773
    50. Shiffman S, Stone AA, Hufford MR. Ecological momentary assessment. Annu Rev Clin Psychol. 2008;4:1-32. [CrossRef] [Medline]
    51. Fuller-Tyszkiewicz M, Dias S, Krug I, Richardson B, Fassnacht D. Motive- and appearance awareness-based explanations for body (dis)satisfaction following exercise in daily life. Br J Health Psychol. Nov 2018;23(4):982-999. [CrossRef] [Medline]
    52. Rogers A, Fuller-Tyszkiewicz M, Lewis V, Krug I, Richardson B. A person-by-situation account of why some people more frequently engage in upward appearance comparison behaviors in everyday life. Behav Ther. Jan 2017;48(1):19-28. [CrossRef] [Medline]
    53. Sonneville KR, Calzo JP, Horton NJ, Haines J, Austin SB, Field AE. Body satisfaction, weight gain and binge eating among overweight adolescent girls. Int J Obes (Lond). Jul 2012;36(7):944-949. [CrossRef] [Medline]
    54. IBM SPSS software. IBM. URL: https://www.ibm.com/spss [Accessed 2025-12-30]
    55. Perneger TV. What’s wrong with Bonferroni adjustments. BMJ. Apr 18, 1998;316(7139):1236-1238. [CrossRef] [Medline]
    56. Bengtsson M. How to plan and perform a qualitative study using content analysis. Nurs Plus Open. 2016;2:8-14. [CrossRef]
    57. Beck JS. Cognitive Behavior Therapy: Basics and Beyond. 2nd ed. Guilford Press; 2011. ISBN: 9781609185060
    58. Hayes SC, Strosahl KD, Wilson KG. Acceptance and Commitment Therapy: The Process and Practice of Mindful Change. 2nd ed. Guilford Press; 2011. ISBN: 9781609189624
    59. Schleider JL, Weisz JR. Little treatments, promising effects? Meta-analysis of single-session interventions for youth psychiatric problems. J Am Acad Child Adolesc Psychiatry. Feb 2017;56(2):107-115. [CrossRef] [Medline]
    60. Fulmer R, Joerin A, Gentile B, Lakerink L, Rauws M. Using psychological artificial intelligence (Tess) to relieve symptoms of depression and anxiety: randomized controlled trial. JMIR Ment Health. Dec 13, 2018;5(4):e64. [CrossRef] [Medline]
    61. Clegg KA. “Feasible but fragile”: an inflection point for artificial intelligence in mental health care. J Med Internet Res. Dec 16, 2025;27:e89202. [CrossRef] [Medline]
    62. Weisz JR, Kuppens S, Ng MY, et al. What five decades of research tells us about the effects of youth psychological therapy: a multilevel meta-analysis and implications for science and practice. Am Psychol. 2017;72(2):79-117. [CrossRef] [Medline]
    63. Heinz MV, Mackin DM, Trudeau BM, et al. Randomized trial of a generative AI chatbot for mental health treatment. NEJM AI. Mar 27, 2025;2(4). [CrossRef]
    64. Wang L, Wan Z, Ni C, et al. Applications and concerns of ChatGPT and other conversational large language models in health care: systematic review. J Med Internet Res. Nov 7, 2024;26:e22769. [CrossRef] [Medline]
    65. Kocaballi AB, Sezgin E, Clark L, et al. Design and evaluation challenges of conversational agents in health care and well-being: selective review study. J Med Internet Res. Nov 15, 2022;24(11):e38525. [CrossRef] [Medline]
    66. Eysenbach G, CONSORT-EHEALTH Group. CONSORT-EHEALTH: improving and standardizing evaluation reports of Web-based and mobile health interventions. J Med Internet Res. Dec 31, 2011;13(4):e126. [CrossRef] [Medline]

    ‎
    ACT: acceptance and commitment therapy
    AI: artificial intelligence
    CALD: culturally and linguistically diverse
    CBT: cognitive behavior therapy
    CONSORT-EHEALTH: Consolidated Standards of Reporting Trials of Electronic and Mobile Health Applications and Online Telehealth
    LGBTIQA+: lesbian, gay, bisexual, transgender, intersex, queer, asexual people, or people otherwise diverse in gender, sexual orientation, and/or innate variations of sex characteristics
    LLM: large language model
    RCT: randomized controlled trial
    SUS: system usability scale

    Edited by Andre Kushniruk, Stephanie Law; submitted 07.Jan.2026; peer-reviewed by Ilaria Pepe, Manuel Alcaraz-Ibanez; final revised version received 23.Jun.2026; accepted 25.Jun.2026; published 20.Jul.2026.

    Copyright

    © Gemma Sharp, Sara Marini, Emily Tam, Hao Hu. Originally published in JMIR Human Factors (https://humanfactors.jmir.org), 20.Jul.2026.

    This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in JMIR Human Factors, is properly cited. The complete bibliographic information, a link to the original publication on https://humanfactors.jmir.org, as well as this copyright and license information must be included.