We followed the PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews) checklist ( to ensure the reliability of our results [52]. To meet the eligibility criteria, the studies had to (1) evaluate an MHCA for efficacy (either in symptom management or user experience), (2) center participants who were the main users of the MHCA, (3) only look at MHCAs that were independent mobile apps or web-based apps with no human involvement, (4) be published in peer-reviewed journals or conference proceedings in English between 2000 and 2024, and, finally, (5) include explicit or implicit references to principles from SAMHSA’s TIC framework [38,46,53]. We did not register a review protocol prior to conducting the study.

We focused on mobile- and web-based MHCAs because these platforms represent the most accessible and widely used forms of digital mental health interventions in everyday contexts. This narrower scope aligns with prior reviews examining CAs for mental health [1,54], allowing for a more nuanced understanding of how trauma-informed design choices are implemented in the tools with which people most frequently engage. We did not include “social robots” or “embodied agents” as search terms but did exclude papers about physical robots, as their design architectures, usage contexts, and modalities differ from app- and web-based chatbots. Some papers used the term “embodiment” to refer to visual depictions or avatars of text- or audio-based CAs; we included these papers.

Articles were excluded if the MHCA was embedded in a preexisting communication platform (eg, Facebook Messenger), ensuring that the system was purpose-built for mental health intervention rather than serving as an ancillary feature. Extended abstracts and posters were excluded because they typically present preliminary or early-stage work that lacks the methodological and analytical depth needed to assess TIC principles. In contrast, we included both commercially available and research prototypes, provided they presented complete or well-documented evaluations relevant to TIC principles. This approach ensured that the review captured both mature systems currently in use and innovative prototypes that may inform future trauma-informed design in mental health technologies (Textbox 1).

To identify relevant articles, we searched Google Scholar, the Association for Computing Machinery (ACM) Digital Library, and PubMed in August 2024 for publications from 2000 to 2024. Our search terms for all 3 databases were as follows: (“mental disorder*” OR “mental health” OR “mood disorder*” OR autism OR “depression” OR “anxiety” OR phobia OR bipolar OR schizophrenia OR affective disorder OR psychosis OR psychotic disorder OR obsessive compulsive disorder OR panic disorder OR post-traumatic stress disorder OR substance abuse OR eating disorder) AND (“conversational agent*” OR “artificial intelligence*” OR “conversational AI*” OR “conversational bot*” OR “CAI*” OR “conversational system*” OR “conversational interface*” OR “smart-bot *” OR “virtual agent *” OR “virtual coach *” OR “avatar*” or “chatbot*” OR “chat bot*” OR “chatterbot*"). These search terms were based on previous studies on MHCAs [1,55-57]. Although suicidality is a critical aspect of mental health, we did not include “suicide” or “suicidality” as explicit search terms. This decision followed prior scoping reviews on MHCAs [1,54], which focused on common diagnostic conditions, such as depression, anxiety, and post-traumatic stress disorder, domains where suicidality frequently co-occurs.

Following the initial retrieval of articles from the databases (N=25,857; ACM Digital Library: n=17,084, 66.07%; Google Scholar: n=220, 0.85%; and PubMed: n=8553, 33.08%), 4 authors (FFN, RP, ER, and KV) independently and manually screened the titles and abstracts to assess eligibility based on predefined criteria. The 4 authors divided the databases year-wise (ie, each was assigned a 5-y span). After this initial screening based on titles and abstracts, out of 25,857, a total of 99 (0.4%) articles were retained. Next, all 5 authors conducted relevancy coding by reviewing the full texts of these 99 articles based on the inclusion criteria. Relevancy coding was documented in Microsoft Excel. Discrepancies were discussed collectively and resolved through consensus. Of the 99 full-text articles, 81 (81.82%) were excluded—1 due to duplication and 80 for not meeting the inclusion criteria (user interaction: 37/80, 46.25%; article type: 5/80, 6.25%; CA type: 20/80, 25%; CA design: 2/80, 2.5%; study type: n=13/80, 16.25%; and no TIC elements: 3/80, 3.75%), resulting in 18 (18/99, 18.18%) studies. To broaden the scope of the review, the authors searched through these 18 papers’ citations to identify additional papers. A total of 62 potentially relevant references were manually screened. By consensus, 20 additional studies were included, resulting in a final sample of 38 papers for review. During screening, we used a broad, inclusive definition of references to the TIC framework that allowed us to explore how its core concepts were interpreted and operationalized across contexts without restricting the search to predefined terminology.

Five authors (FFN, FK, RP, ER, and KV) coded the data in Excel for a broad array of information on each publication in the corpus; publications were randomly assigned among authors. They extracted basic information per publication, including study methodology and outcomes and characteristics of the CA intervention, aligned with the PRISMA-ScR structure and previous studies [1]. Apart from the initial characterization of the studies, the analysis followed an abductive, theory-informed approach [58], combining deductive sensitization to SAMHSA’s TIC principles with openness to inductively derived subthemes reflecting how these principles were operationalized across studies. As none of the papers explicitly referenced or applied SAMHSA’s TIC framework, identifying the 6 TIC principles required careful interpretive analysis. Descriptive and interpretive analyses of all extracted TIC-related data were conducted by the first and last authors, who iteratively revisited the data throughout the writing process to ensure that nuances in how TIC-related ideas were represented were accurately identified and not overlooked. To ensure methodological rigor and reduce subjectivity, the authors used a multistep, consensus-driven coding process. Discrepancies were systematically discussed and resolved through consensus, followed by iterative recoding to refine consistency and reliability. While this approach strengthened the credibility and reproducibility of our thematic interpretations, no qualitative work is neutral, and all interpretation was shaped by authors’ professional experiences and positionalities, as detailed in the next section.

Embedded within a large health system, the interdisciplinary research team includes members with expertise in HCI (FK and FFN), social computing (FK and FFN), public and mental health domain (KV, RP, and FFN), health informatics (FFN, RP, and FK), and user experience research and design (ER, FFN, and FK). The team’s work is informed by ongoing collaboration with clinical providers and experience in engaging with mental health–related data, digital interventions, and sociotechnical systems in health care settings.

Clinical input from a licensed provider (a clinical psychologist from the same health system with expertise in trauma and TIC) was sought during the design of the study. Some team members have participated in trauma-informed technology design training within and outside of their health institution. However, our perspective is primarily grounded in applied informatics research and health care practice, shaped by close collaboration with clinicians and work in mental health settings. The team also includes researchers from both Western and non-Western backgrounds.

The TIC framework, as defined by SAMHSA, provides a foundational approach for recognizing and responding to the impact of trauma across health care and other service systems [38]. Within this framework, trauma is defined as a combination of experiences that an individual perceives as harmful or life-threatening and the lasting adverse effects of those experiences on the individual’s functioning and well-being [38,46]. Rooted in research and expert consensus, the TIC framework is guided by six interconnected principles: (1) safety, ensuring physical and emotional safety in environments and interpersonal interactions; (2) trustworthiness and transparency, building trust through transparent, consistent, respectful, and fair communication and decision-making; (3) collaboration and mutuality, promoting partnership and reducing power imbalances between individuals, whether staff or clients; (4) empowerment, voice, and choice, recognizing and strengthening individuals’ existing capacities, voices, and experiences; (5) peer support, valuing and incorporating the perspectives and support of those with lived experiences of trauma; and (6) cultural, historical, and gender issues, being responsive to cultural, racial, historical, and gender-based contexts that shape individuals’ experiences [38,53]. These principles are designed to promote environments that acknowledge trauma’s pervasive effects, recognize its signs and symptoms, integrate this understanding into practice, and actively seek to prevent retraumatization [46]. We apply the TIC framework as a critical lens for evaluating mental health chatbots (MHCAs), as these digital tools often interact with users during moments of psychological vulnerability, and their design choices can either mitigate or exacerbate distress.

Checklist 1 represents the PRISMA-ScR checklist we followed. The study selection process is summarized in Figure 1. A total of 18 (47.4%) publications were identified through the initial database search (PubMed: 9/38, 23.7% [12,13,59-65]; the ACM digital library: 7/38, 18.4% [66-72]; and Google Scholar: 2/38, 5.2% [73,74]; Tables 1 and 2). The remaining 20 (52.6%) publications were found as citations of these 18 sources [75-94]. Overall, 30 out of 38 (78.9%) were journal articles, with the most frequent venues being the Journal of Medical Internet Research and affiliated journals (15/30, 50% articles), followed by Frontiers in Digital Health (4/30, 13.33% articles); 8 out of 38 (21.1%) were conference proceedings. Most (28/38, 73.7%) studies were published in 2020 or later, reflecting a recent surge in the domain. The United States was the most common study location (13/38, 34.2% studies), whereas 12 (31.6%) out of 38 articles involved users based in various European countries. While most studies’ participants were from a single country, 4 (10.5%) studies included participants from multiple countries [66,76,83,84], signaling some movement toward globally inclusive development.

A total of 18 (47.4%) studies included in this scoping review were randomized trials (randomized controlled trials: n=17, 94.4%), with 4 (22.2%) identified as preliminary or pilot studies. Sample sizes for randomized controlled trials ranged from 30 to 700 (mean 140.35, SD 153.99; median 107, IQR 58-148). Additionally, 5 (13.2%) studies used other experimental designs, including between-subject studies [66,76], single-arm pre-post intervention studies [74,83], and nonrandomized prospective studies [94]. Eleven (28.9%) publications were based on user studies and included mixed methods (n=6, 54.5%) [13,67,73,78,79,82], qualitative approaches (n=3, 27.3%) [60,68,77], and quantitative approaches (n=2, 18.2%) [70,75]. Observational analyses of commercial app data (5/38, 13.2%) [13,65,78,88,89] featured the largest sample sizes of all study types (mean 2334.8, SD 1775.8; median 2194, IQR 667-4073). Finally, 1 (2.6%) study comprised a quantitative survival analysis using data originally collected from a separate clinical trial [62].

Overall, participant samples in the reviewed studies skewed younger (mean age 32.52, SD 14.6 y) and female (male participants: mean 34.92%, SD 18.64%). Twenty-nine (76.3%) publications involved only nonclinical samples, whereas 7 (18.4%) included clinical samples. Two (5.2%) studies recruited both clinical and nonclinical samples for comparison [67,83]. Depression and anxiety were the most frequently addressed concerns across the studies included [12,13,59,61,62,67,70,78,80,83,86]. Summaries of publications can be found in Table 1; more information is available in Multimedia Appendix 1.

Most MHCAs were reported in a single publication, although a few appeared in multiple studies (eg, Woebot was included in 6 studies, and Wysa was included in 5 studies). While most publications evaluated a single MHCA, some did multiple MHCAs [73,90,93]. Thus, 28 distinct MHCAs were represented across the corpus. Seventeen (60.7%) of 28 MHCAs were described as prototypes at the time of the study, and 9 (32.1%) were commercially available. However, close to half (17/38, 44.7%) of the publications in the corpus focused on commercial MHCAs. The most popular MHCAs were Woebot [12,68,70,73,74,87] and Wysa [13,62,73,78,88]. Text (including emojis) or multiple-choice options were the most common input and output modalities, with 13 (13/28, 46.4%) MHCAs including other modalities such as audio or video. Sixteen (16/28, 57.1%) MHCAs used rule-based functionality for conversation logic, 1 (1/28, 3.6%) was fully generative AI, and 9 (9/28, 32.1%) incorporated both. Eighteen (18/28, 64.3%) MHCAs were described to have a visual avatar (eg, realistic animated face [84] or nonhuman abstract character [77]). Fourteen (14/28, 50%) MHCAs were described as having a form of crisis intervention. While other MHCAs could have this feature, it was not directly mentioned in the articles.

One component of design frequently absent from included articles was the extent to which the MHCA versus the user guided the interaction, as well as examples of representative interactions. Out of 28 MHCAs included in this study, 26 (92.9%) delivered interventions to improve mental health; one (3.6%) focused on diagnosis [82], and another (3.6%) focused on hospital discharge counseling [80]. Mental health improvement interventions were most often based on CBT (14/26, 53.8%), psychoeducation (8/26, 30.8%), mindfulness (4/26, 15.4%), positive psychology (3/26, 11.5%), motivational interviewing (2/26, 7.7%), self-care or self-help (3/26, 11.5%), acceptance and commitment therapy (2/26, 7.7%), gratitude (2/26, 7.7%), and mood tracking (2/26, 7.7%). Eleven (42.3%) of 26 CAs that provided mental health interventions were multimodal, involving 2 or more different approaches. Summaries of MHCA interventions are presented in Table 2, with more information on MHCAs provided in Multimedia Appendix 2.

Whereas the previous sections document how TIC principles appeared across MHCA research, this section presents concrete design recommendations. As most reviewed papers did not include formal design recommendation sections, we extracted actionable guidance wherever authors reflected on design implications, including system descriptions, findings, and discussions. As design recommendations can intersect with multiple TIC principles, our classifications are intentionally flexible rather than rigid and should not be interpreted as mutually exclusive. Textbox 2 presents a summary, while Multimedia Appendix 4 breaks overarching themes down by publication, categorizing each reference according to which explicit principle (ie, Table 3) or implicit concept (ie, Table 4) it is connected to.

A core requirement across studies was building safe interaction spaces, where users felt unjudged and confident in how their data and disclosures were managed. Many studies emphasized that anonymity, whether through no registration models or optional pseudonyms, reduced fear of judgment and facilitated self-disclosure [60,68,78,83,88]. Conversely, using names was also found to build trust between the chatbot and the user [71,76], although it could raise privacy concerns [76]. Some studies bridged these concerns by recommending MHCA apps be password protected [89,90].

Similarly, transparent communication around data storage practices (eg, what is stored, for how long, and by whom) was tied to perceptions of trustworthiness [67,68,83,87]. Users were more willing to disclose information when they knew it was not archived indefinitely or shared across platforms [68,76]. These insights mapped onto the digital safety–related themes in Table 4. While secure data practices established the foundation of trustworthiness, usability, and availability further determined whether users experienced the MHCA as a dependable source of support. Several studies reiterated that chatbots should be accessible 24/7, enabling users to seek support from home or other safe spaces, reducing feelings of helplessness, and ensuring timely assistance [60,64,68]. Furthermore, interfaces should be intuitive and easy to navigate, addressing usability challenges [64,73] and providing usable, stigma-free emotional support [74].

Physical and emotional safety were further addressed when MHCAs incorporated crisis intervention mechanisms, such as save our souls (SOS) buttons, trigger-word detection, access to emergency helplines, and tailored resources for suicidal or self-harm–related disclosures [59,63,64,73,74,76,77,83,86,92]. Transparent statements clarifying that the CA is not a crisis service also supported trustworthiness by setting clear boundaries [74,87]. Together, these practices mirror themes related to safety, trustworthiness, and transparency outlined in Table 4.

Users appreciated empathic, relational behaviors from MHCAs but still wanted clarity about their nonhuman identity. MHCA designs communicated their nonhuman nature, purpose, capabilities, and limitations upfront through robotic names or avatars, onboarding disclosures (eg, “digital coach vs therapist”), or welcome messages [68,74,77,79,83,85,87,89]. This transparency promoted user autonomy, confidence, and motivation, as users could engage with the system with clear expectations of its role and limitations [76]. Conversely, artificial backstories by chatbots, exaggerated friendliness, or attempts to appear human were often perceived as deceptive and diminished trust [66,84], highlighting the importance of neutral, “robotic” visuals, such as cartoon avatars [79]. While clarifying the MHCA’s nonhuman identity set the foundation for trustworthy engagement, users also relied on ongoing transparency within the interaction, including about data use and crisis handling. Providing clear instructions, menu navigation, and orientations helped users navigate the system efficiently, prevented misunderstandings, and supported self-efficacy and confidence in interacting with the MHCA [59,60,77,79,83,90,93]. One study also suggested that the user be able to view and modify what the MHCA had learned about the user to improve transparency [71]. Overall, transparency about data use and crisis management boundaries improved predictability, engagement, and trustworthiness, reflecting concepts of digital safety and responsible use [74,76,83].

Other relational design elements, such as checking in regularly [62,70], engaging in active listening [66,70], or expressing sympathy [74,76,85], supported positive emotional and psychological experiences of use. These features promoted emotional safety and enhanced user confidence, motivation, and comfort [68,74,85]. Well-designed avatars could further strengthen rapport and comfort, as user preferences modulated the effects of customized anthropomorphic versus neutral representations [72,79,82,84].

Designing for trustworthiness in MHCAs was closely tied to the system’s accuracy, reliability, and consistency. Some studies suggested that it is critical to provide clinically validated content where the materials are predefined and professionally reviewed, instead of generative AI outputs, to ensure all responses are validated, clinically safe, trustworthy, reliable, and of high quality [60,68,76,78,86]. MHCAs should clearly indicate whether individual pieces of content have been verified to enhance credibility [60,68,76]. Additionally, MHCAs should explain why specific interventions are selected and identify the source or author of self-care modules so users can properly understand and evaluate recommendations [76,77].

Receiving the right kind of help, accurate interventions, and contextually appropriate responses was critical for maintaining confidence in the system [60,64,68,70,76,78,91]. Technical issues, including glitches, slow loading, or failure to respond to unexpected user inputs, could undermine trust and engagement [59-61,64]. Similarly, inconsistencies in intervention delivery or errors in translating content (eg, English to Spanish) reduced perceived reliability [60,91]. Alongside reliability, adaptive and nonrepetitive content sustained user trust and perceived usefulness. Users valued responses that were tailored to their experiences, were not repetitive, and adapted dynamically based on feedback or prior interactions [13,60,61,64,70,76,81,84]. Incorporating user-specific language and ensuring real-time responses made interactions more understandable, relatable, and trustworthy [87,89]. Conversely, rigid or repetitive patterns or failure to remember prior conversations decreased trust and disrupted collaboration [59,64,86].

Supporting user autonomy and empowerment through flexibility and personalization was another central theme. Several studies discussed the importance of users taking control of MHCA interactions and settings. For instance, allowing users to lead conversations [78] or choose areas of focus within an app, such as Wysa [73], increased autonomy and supported therapeutic alliance [78]. Scrolling back through past interactions [80,90], opting in or out of reminders [63,92], and determining the timing of check-ins [62] gave users control, promoting confidence and ownership over their healing.

Embedded customization and flexibility were also common themes across studies. Allowing users to personalize the personas of therapeutic agents [72], select intervention components [85], or modify avatars to reflect their preferences [71,75] created a sense of control, comfort, and trust. Similarly, giving users the option to translate speech to text [84], download or review chatbot content [64], and navigate flexibly through the system [60] promoted self-efficacy [66]. Systems that supported multiple modes of input and output (eg, text, free text, audio, video, clickable buttons, emojis, graphical interfaces, and PDF or figure uploads) further empowered users to engage in ways that matched their preferences and needs [59,60,64,84,85,88,89]. Alternatively, providing predefined options alongside free-text input, with clear protocols when input is not understood, balanced flexibility with safety [59,60,66]. Beyond MHCA settings and interface, designing shared goal setting, check-ins, and accountability mechanisms operationalized user empowerment and collaboration. For example, Woebot and similar agents allow users to set goals, track progress, and engage in regular check-ins [12,62,74,92]. Daily or weekly follow-ups and reminders help maintain engagement, provide structure, and signal that user input is valued, thereby indirectly reinforcing both collaboration and empowerment [84-86] alongside regular reports summarizing user mood, progress, or activity [12,13,62,73,79,85].

Design recommendations related to sensitivity to cultural, historical, and gender issues included accessibility and responsiveness to user preferences, language, and characteristics. Users requested accurate multilingual support and content that reflects their lived realities, including gender, race, and age [60,72,73,77,83,91]. Implementing features such as a gender-nonspecific avatar [77], agents that mirrored the user’s race [72], youth-appropriate language, nontriggering naming options, and localization affected inclusivity, trustworthiness, and perceptions of safety [66,71,79]. Accessibility included specifics of basic, usable interface design, with users valuing readable fonts and clear icons [67], as well as nonverbal or text-only interaction options in noisy or vulnerable environments [80,84]. Context-aware behaviors, such as avoiding inappropriate suggestions when users are distressed, further supported emotionally responsive interactions, indirectly reflecting the idea of sensitivity [81].

Although SAMHSA’s TIC framework was originally developed for clinical contexts, it has been increasingly adapted in nonclinical domains such as educational institutions [96] and workplaces [97], and there is growing interest in applying TIC within HCI research [43,44,95]. Despite this uptake, the TIC framework has yet to be systematically applied to MHCAs, which is a relevant opportunity, given these systems’ interaction with users in vulnerable psychological states. This scoping review represents an initial effort to bridge clinical and technological domains, connecting research and practice across the heterogeneous MHCA literature. We chose to expand our analysis beyond mentions of explicit terminology to reflect the intrinsic subjectivity and interconnectedness of the TIC framework and to more comprehensively identify and summarize how scholarship to date has operationalized TIC-related ideas. In this section, we highlight key opportunities for future work to leverage the TIC framework in MHCA development, informed by the range of TIC-related language, constructs, metrics, and design recommendations identified in our review.

Our scoping review of 38 publications reveals a nuanced and varied but fragmented landscape of MHCAs that implicitly reflects concepts aligned with the TIC framework. Many studies addressed safety (eg, anonymity, secure data practices, positive emotional experience, and crisis infrastructures); trustworthiness and transparency (eg, accuracy, reliability, signposting, and nonhuman identity disclosure); empowerment, voice, and choice (eg, control over settings, personalization, and flexibility); and collaboration and mutuality (eg, therapeutic alliance, check-ins, and accountability). Other TIC principles receive far more limited attention: peer support and sensitivity to cultural, historical, and gender issues were notably absent, not only in design but also in evaluation and theoretical foundations. Furthermore, while we identified design recommendations across studies that reflected TIC principles (eg, ensuring content quality, providing disclosure about MHCAs, offering control and customization), these recommendations were at times vague, emphasized different domains, or even presented contradictory guidance (eg, anonymity vs identity and customization, human-like vs clearly robotic). This reflects that developers lack standardized guidelines, checklists, or design matrices to systematically implement TIC principles. Finally, while it is encouraging that TIC-related constructs such as accuracy, consistency, security, therapeutic alliance, goal setting, and mutual respect are increasingly measured using validated quantitative scales (eg, WAI-SR, CUQ, and SUS), in many instances, TIC-related elements are relegated to background, discussion, and supplementary qualitative findings and were not central to studies’ or MHCAs’ purpose, approach, or design.

While large language models (LLMs) are gaining traction in mental health applications [98-100], most MHCAs in included publications were completely rule-based, inviting discussions about safety. The use of predefined, structured interactions over open-ended dialogue was often promoted in the corpus, highlighting current limitations in natural language processing capabilities, computational resources, and the need to minimize clinical risk from inappropriate or incoherent responses. Other prior work underscores that LLM-based CAs are not yet mature or reliable enough for psychologically and physically safe implementation in high-stakes, trauma-sensitive conversations [101,102]. From a trauma-informed perspective, recognizing acute vulnerability and preventing harm are foundational considerations.

Although some studies referenced crisis escalation or support mechanisms, this review did not systematically sample research explicitly addressing suicide risk. This limited our ability to evaluate how comprehensively MHCAs address safety and crisis vulnerability, particularly given that agents designed for depression and anxiety may interact with populations for whom suicidality is clinically relevant. Despite this, our review still surfaced a critical safety gap as half of the MHCAs in the corpus did not describe crisis intervention mechanisms or protocols. Overall, discussions about safety were scattered across paper sections, and a unified sense of real or perceived safety was rarely evaluated, raising concerns about how psychological and physical safety are truly ensured during MHCA interactions and echoing another recent systematic review that identified safety as underevidenced in many MHCA deployments [10]. Addressing these gaps may support the development of measurable, standardized requirements that promote both clinical effectiveness (eg, when using measures such as the Patient Health Questionnaire-9) and user safety, particularly for trauma-affected populations.

Crucially, too, our corpus included little engagement with cultural, historical, and gender issues, which influence many forms of traumatization and which are intended in the TIC framework to underlie and inform all parts of clinical and organizational practice [53]. Accessibility and usability, common focuses in included studies, can be implicitly connected to cultural, historical, and gender issues but are limited by the demographics of participants whose usability and access needs are represented. The influence of systemic and historical power dynamics on users’ experiences with MHCAs was largely unexplored, beyond relatively narrow references to demographic- or diagnosis-specific needs [62,79,88]. It is possible that researchers addressed these issues in practice in researcher-participant dynamics (eg, through institutional review board interactions) but did not report them in publications or did not incorporate them in MHCAs beyond study design ethics.

This represents a significant missed opportunity. Explicitly incorporating cultural, gender, and historical factors in MHCA design, research, and user experience could enhance intervention customization and user satisfaction [66,80], while also addressing pressing ethical challenges, such as user safety [27,103,104], LLM-related bias [27,105,106], access and literacy gaps [10,107,108], and the tools’ ability to handle nuance and complexity [103,104,109]. The issue of rule-based versus LLM-based interventions is also extremely salient for addressing sensitivity to cultural, gender, and historical factors. Depending on how they are trained or fine-tuned (which may frequently be entirely opaque), different LLMs can show differential orientations toward specific cultural values, displaying biases and reinforcing stereotypes [106] or struggling to appropriately handle race [105]. Algumaei et al [10] echoed this in their 2025 systematic review; they found that effective localization and cultural adaptability in MHCAs varied by implementation and that deficits in accessibility and inclusivity (especially for low-resource settings and Western-oriented systems) persist. Overall, the relative absence of potentially unsafe or biased LLM-based MHCAs, of written detail on MHCA interventions, and of historical sensitivity in the corpus is a critical gap where the TIC framework may provide guidance.

Finally, peer support was nearly absent from the corpus. For the 3 papers that mentioned it, recommendations were limited, with some suggestions that MHCAs encourage seeking out in-person support from friends and family [73,77,79] and that it might include a “seek support group” option [77]. Furthermore, it is theoretically unclear whether users perceive relationships with MHCAs as peer support and equally unclear how MHCA designers and researchers intend this relationship to be characterized. The fact that papers in our corpus variably used therapeutic alliance, friendship, or user tool to characterize the user-MHCA relationship, as well as the conflicting recommendations for human-like versus clearly robotic MHCA, warrants further attention. Scholarly work has criticized positioning MHCAs as therapists [27,28], but other relational dynamics, including MHCAs as peers, are less well-explored.

A key finding of our analysis was that the field lacks a shared, digital CA-specific framework to justify and evaluate MHCA design and performance through a lens that is both trauma-informed and based in best clinical practice. The variety of terms and concepts (Table 3 and Table 4) and metrics (Multimedia Appendix 5) we identified as related to the TIC framework, in the absence of citing it directly or completely, results in inconsistent definitions, even when publications use the same terminology (ie, “trust” in the context of the Trust in Automation scale may be different than “trust” in the context of therapeutic alliance). This complicates researchers’ ability to identify patterns in quantitative outcomes and design recommendations, fragmenting interrelated ideas across CA-related domains and disciplines.

While we noted concepts related to TIC principles to be assessed completely or partially by an array of quantitative scales, this may also lead to redundancy, as multiple tools assess similar ideas but use differing theoretical framings or language. For example, the CUQ (used by Boyd et al [67]), designed to evaluate CAs, includes a question implicitly measuring its trustworthiness and safety: “Chatbot responses were useful, appropriate, and informative” [110]. The Client Satisfaction Questionnaire, as it was cited in a study by Liu et al [86], is validated for residential substance abuse treatment and includes a comparable question about usefulness: “Have the services you received helped you deal more effectively with your problem?” [111]. However, both questionnaires also include questions that do not overlap and were intended for different contexts, making overall CUQ or Client Satisfaction Questionnaire scores, or even answers to similar questions, incomparable. Without a holistic framework or metrics, it is difficult to build a clinically based, evaluable standard for MHCAs’ therapeutic experience that goes beyond clinical symptom outcome scores such as the Patient Health Questionnaire-9 or Generalized Anxiety Disorder-7 questionnaire.

However, it is not necessary to find one-to-one mappings of TIC principles. While fragmented, the interdisciplinarity of this topic area and the heterogeneous terminology we identified also illustrate many opportunities to adapt, combine, and validate constructs and measurements from a variety of fields, such as psychotherapy (eg, working alliance [112] and self-efficacy [113]), user experience design (eg, accessibility [114] and usability [115]), HCI (eg, Trust in Automation scale [76] and privacy and security [116]), and study of interpersonal relationships (eg, self-disclosure [117] and relational flourishing [118]). The body of research on explainable AI may provide use for increasing trustworthiness and transparency [119]. Drawing from critical studies, such as feminist [120], decolonial [121], and disability justice [122,123] approaches, may be useful for identifying relevant cultural, historical, and gender issues and beginning to address them. Scott et al [124] note that approaches to accountability and repair (eg, from restorative justice) echo TIC’s focus on those harmed; when MHCAs breach safety, trust, transparency, or fail to properly consider cultural, historical, and gender issues, these may also be highly relevant. Given the TIC framework’s American origin and interest in inclusivity, as well as well-known inequities in who MHCAs are developed with and for [10,107,108], future integrative development should include cultural, historical, and gender considerations that affect diverse MHCA users, not only those who are easy to reach or accommodate. In our corpus, age [68,70,73,77,84,91] and diagnoses and symptom patterns [62-64,75,79,88,92] were common sampling mechanisms. However, diagnosis and screening tools are often functions of culture and access [125-128], and salient demographic concerns in AI development also include gender [129], disability status [130], and more.

While the precise, practical, and preferred operationalization of the TIC framework and all its principles in the MHCA domain remains a question for future work, this scoping review has identified that an array of TIC-related ideas, regardless of their feasibility, are already present in the current academic landscape. The interconnectedness of the concepts we identified (eg, working alliance reflecting trustworthiness and transparency and collaboration and mutuality; customization reflecting empowerment, voice, and choice and collaboration and mutuality) suggests that going a step further to include more or all of TIC’s interrelated principles will likely be consolidating and descriptive; addressing one principle often directly or indirectly reinforces others. A more intentional synthesis, as well as filling the gaps noted earlier, under the umbrella of the TIC or trauma-informed computing framework, could help unify the inconsistencies we identified as well as inject empirical, clinical, and organizational best practices into the design of these implicitly relational digital health tools.

A key limitation of our scoping review is that no paper explicitly referenced the TIC framework or consistently defined its principles in the context of digital interventions. As a result, we had to infer alignment with TIC based on interpretation rather than author intent, introducing some subjectivity and internal biases in the results based on the authors’ specific backgrounds (ie, in HCI, public health, and user experience and not in applied TIC). Additionally, as only one study specifically focused on patients with trauma and many deliberately avoided including participants with severe mental health concerns, our ability to surface more nuanced design implications specific to trauma-affected users was constrained. This sampling tendency among publications also resulted in a marked absence of suicidality as a concern in MHCA and study design. We note that each aspect of TIC is highly complex; while our analysis intentionally adopted a holistic, scoping lens, each principle warrants future in-depth engagement. In particular, equity- and historical trauma–centered approaches may lead to different and more in-depth conclusions with regard to the differential impacts of MHCA design and deployment.

Furthermore, not all CAs used for mental health support are labeled as MHCAs, which may have excluded studies from this academic corpus. Many publications provided minimal detail on chatbot interfaces, workflows, or example interactions, limiting our ability to assess their alignment with TIC principles or determine whether systems were rule-based or AI-driven. Most MHCAs in our corpus were also prototypes, offering limited visibility into deployed or widely used systems. Finally, our review was limited to English-language publications and focused on the US-based SAMHSA TIC framework, which may have excluded relevant international or culturally specific approaches. Future research could broaden this scope by including gray literature, commercial documentation, and non-English sources, as well as considering alternative or complementary frameworks that reflect diverse global perspectives on trauma-informed design.

The aim of this scoping review was to identify how the TIC framework is currently being adopted in the evaluation and design of MCHAs in academic literature. While no prior studies explicitly applied the TIC framework, we identified numerous instances where TIC-related concepts were referenced. Most commonly, these references were to trustworthiness and transparency; empowerment, voice, and choice; collaboration and mutuality; and safety, with peer support and cultural, historic, and gender issues largely overlooked. These principles were typically described in intervention design or discussion sections rather than rigorously evaluated. We also observed emerging design trends that support trauma-informed approaches, including customizability, consistency, flexibility, accuracy, and positive emotional expression. Collectively, these findings highlight opportunities to develop holistic, clinically informed, and trauma-informed design guidelines, metrics, and evaluation methods for MHCAs, providing a foundation for future interdisciplinary research in this space.