How Can We Make Digital Mental Health Tools Work for Those Who Need Them Most?

Image by DALL E-3

Digital health tools often overlook those who need it most: people with serious mental illness, marginalized communities, and non-English speakers. If equity is the goal, these groups must be the starting point of design, not an afterthought.

A mother in a refugee camp taps open a mental health app. It promises support but the language is English, the visuals buffer slowly, and the meaning is lost before she even begins.

A teenager in East Asia scrolls through a chatbot. It asks her to “be assertive” and “prioritize herself.” But in her world, those ideas clash with how love, duty, and respect are expressed.

An older man living alone in rural Canada tries a wellness tool. The colours are bright. The icons animate. But it’s all noise to him. He can’t find the words he needs behind the emojis, badges, and busy screens.

A young woman in a crowded home gets a reminder to check in. But she shares her phone with two siblings. The data is low. The prompt flashes and disappears before she can safely answer.

These aren’t edge cases. They’re the silent norm. Most digital mental health tools weren’t built for the people who need them most. They were built for convenience then labeled as care. And in doing so, they replicate the same inequities they claim to fix.

Digital mental health tools, from therapy apps and chatbots to online counselling platforms, hold tremendous promise. In a world where nearly one billion people are living with a mental disorder, digital interventions could bridge the treatment gap in mental health care. The World Health Organization (WHO) estimates that 76–85% of people with mental health conditions in low- and middle-income countries receive no treatment at all. Even in wealthier nations, services fall short: only about one-third of people with depression get adequate care in high-income countries, and as few as 3% do in low-income countries. Meanwhile, the global supply of mental health professionals is extremely limited (a global median of just 13 mental health workers per 100,000 people) and unevenly distributed (high-income countries enjoying over 40 times the workforce available in poorer regions). This shortage, coupled with stigma and long waitlists, leaves millions without support.

This is the digital divide in mental health: global in ambition, local in failure.

The COVID-19 pandemic only intensified the need. Rates of anxiety and depression surged by roughly 25% worldwide in the first year of the pandemic, further widening the gap between demand and available care. Digital platforms raced to fill this void, offering self-guided therapy programs, mood trackers, AI chatbots, and tele-therapy. Many have attracted millions of users, drawn by the promise of on-demand, affordable help that can reach people in remote or underserved areas where traditional services are scarce.

Too often, however, these digital tools are designed without enough input from the people who will actually use them. As a result, their real-world impact has fallen short of the hype. Many apps see high download numbers but low long-term engagement. Users try them once or twice and drop off, finding them irrelevant, burdensome, or not addressing their true needs (see my articles “Why Do Mental Health Apps Struggle to Retain Users” and “Why We’re Ghosting Mental Health Apps and Finding Solace in AI”). Others may work well for a general audience but fail to serve those with the most serious or complex conditions.

The people with the greatest needs, such as those with severe mental illness, those in marginalized or low-resource communities, or non-English speaking populations, are often an afterthought in digital health design.

If we truly want these innovations to help those who need them most, we must fundamentally rethink our approach. This means:

• Designing with end-users from the start.
• Adapting solutions to local cultures and contexts.
• Rigorously evaluating effectiveness, and
• Upholding the highest standards of privacy, safety, and ethics.

Clinicians, developers, and patients need to work together to turn digital promise into real progress in mental health care. What follows is my exploration of key strategies informed by recent research and case studies to make digital mental health tools more adaptable, accessible, and effective for the people who need them the most.

Build With Users From the Start

One significant reason digital mental health interventions often miss the mark is a lack of meaningful user involvement in their creation. Traditionally, apps or online programs have been designed by developers or researchers in a top-down manner. The result can be polished technology that technically “works,” but doesn’t resonate with those it aims to serve.

Engaging end-users in the design process, through human-centered design, participatory design, and “design thinking” methodologies, is critical to building tools that people will actually use and benefit from.

Design thinking, in particular, has gained traction as a strategy to refocus digital health innovation around the human perspective. This approach is most often used in fields like consumer tech and product design and emphasizes empathy with users, iterative prototyping, and testing solutions in real-world contexts. In the realm of youth mental health, for example, researchers Hanneke Scholten and Isabela Granic argue that applying design thinking can address key shortcomings of current digital interventions such as poor engagement, lack of personalization, and low real-world efficacy. In a 2019 article, they note that embracing a design thinking mindset would allow developers to tap into an entirely new toolbox of practices to create more engaging, tailor-made interventions.

In practice, design thinking means starting by deeply understanding the daily lives, preferences, and challenges of the people we aim to help, before a single line of code is written. For instance, the development team at the Games for Emotional and Mental Health (GEMH) Lab in the Netherlands works directly with teens to co-create mental health video games. They hold genuine conversations with young people about their needs and involve them throughout the development process (a technique known as participatory design). One outcome of this approach is *MindLight*, a therapeutic video game for children with anxiety that uses gaming, neuro- and bio-feedback to teach coping skills. By iteratively testing prototypes with youth and integrating their feedback, the designers ensured the end product was not only clinically sound but also fun and appealing to its intended audience.

Another example comes from the global mental health sphere. The SUPER Project (Successful User Participation Examples and Recommendations) in digital mental health explicitly set out to answer how we can build better tools by engaging end users and clinicians in every step of creating and implementing mental health technology. The project piloted a transnational approach: a Dutch app for autism stress management was adapted with input from Danish users, and conversely a Danish youth stress app was modified for Dutch users, each time incorporating local user feedback at every stage.

One-size-fits-all solutions rarely work in mental health. We must co-create and co-adapt tools with the very people who will use them, in their specific cultural and care contexts.

The early insights from initiatives like SUPER are promising. According to project leads, digital mental health solutions become more adaptable, accessible, and effective when local expertise is included and iterative design is embraced. Even simple changes based on user input like adjusting an app’s language to fit local slang, modifying the timing of reminders to suit users’ daily routines, or adding features users requested, can significantly improve engagement and outcomes. This philosophy echoes a broader shift in healthcare toward patient-centered and community-led innovation.

Involving users cultivates a sense of shared ownership and trust. People are more likely to trust and stick with a tool that they or their peers had a hand in shaping.

For those of us building or recommending these tools, the takeaway is clear: we need to bring patients and front-line providers to the table from day one. This means:

• Conduct focus groups and interviews to learn what barriers your target users face.
• Co-design features and interface elements with diverse users, especially those from populations at highest risk or with unique needs (e.g. youths, seniors, people with serious mental illness, minorities).
• Test early and often in real-world settings. A prototype only truly successful when it seamlessly fits into users’ lives.

To truly fulfill the promise of effective, interactive online therapy, we need to build with the people we aim to serve, not for them.

Tailor For Culture and Context

Mental health is deeply influenced by cultural norms, language, and local realities and an app or bot that works well in one setting may flop in another if these factors are ignored. Thus, effective digital mental health tools must account for cultural and contextual differences across communities. For example, a cognitive-behavioural self-help app developed in English-speaking North America might not readily translate to helping a rural community in South Asia without adapting the content, language, metaphors, and even delivery mode to fit the target population’s context.

Recognizing this, leaders in global mental health have called for “glocalization” of digital interventions (i.e. integrating global knowledge and technology with local tailoring). This approach recognizes that while digital mental health interventions (DMHIs) offer scalable and accessible solutions, their effectiveness depends on being adapted to fit the cultural, linguistic, and contextual needs of local populations. The WHO, for example, has developed and tested digital therapeutics that emphasize cultural adaptation. A notable case is Step-by-Step, a guided digital behavioural program for depression. It was co-developed with local partners in the Middle East and customized for Syrian refugees’ context and language. In a controlled trial among Syrian refugees in Lebanon, the Step-by-Step intervention significantly reduced depression and anxiety compared to standard care. The tool used simple illustrated stories and audio in colloquial Arabic, and provided weekly support calls from trained lay helpers, aligning with the users’ literacy levels and cultural preferences. These adaptations paid off as refugees who used the digital program showed significantly better functioning and lower distress than those who did not, with benefits sustained at follow-up. Importantly, the success was replicated among Lebanese and other groups living in the same region.

The WHO champions Step-by-Step as a model that can be scaled to other displaced populations with internet access, precisely because it was designed with local culture and feasibility in mind.

Cultural tailoring isn’t only about language translation, it extends to how mental health problems are framed and what solutions are acceptable. In some cultures, formal therapy is stigmatized, but people might seek help via clergy, community leaders, or traditional practices. Hence, a digital tool will be more effective if it complements these local help-seeking behaviours and doesn’t inadvertently clash with them. For instance, a mental health chatbot intended for use in India might incorporate culturally familiar examples and respect local communication styles (e.g., being less direct about personal issues at first, or using relatable narratives from Indian family life). Likewise, an app for teen mental health in East Asia might need to consider different family dynamics and attitudes toward mental illness than an app for Western teens.

Cultural context matters not only for translation but for core design decisions: what motivates users, what their daily routines look like, whom they trust, and what technology they have access to.

Another important aspect of context is accessibility and platform. In low-resource settings, leveraging technologies like SMS text messaging might be crucial for reaching those most in need. For example, some successful global mental health interventions have used SMS prompts for supporting patients (akin to cognitive behavioural therapy exercises delivered via text) because SMS is universally available on even the simplest phones. This is context-aware design. It means meeting users where they are, literally and figuratively. As examples, if your target users live in a refugee camp with limited connectivity, a lightweight, text-based intervention might work better than an app with large multimedia files. Or if your users are older adults who aren’t tech-savvy, the interface must be extra-simple and perhaps supplemented by telephone (human) support.

Digital mental health solutions must be flexible enough to accommodate diverse environments and cultures. By incorporating these factors, we avoid the pitfall of creating a tool that is technically effective but culturally irrelevant.

We need to ask, together, as teams designing for mental health across cultures:

• What languages does the audience speak?
• What are their shared beliefs about mental health?
• Do users have privacy when using a phone?
• What devices and bandwidth do they have?

Insist on Evidence, Not Just Engagement

For clinicians, a key concern is whether these digital tools actually work. Unlike prescription medications or medical devices, most mental health apps launch direct-to-consumer with minimal oversight or requirement to prove their efficacy. The result is a Wild West of digital health, where apps might claim to “treat” depression or anxiety without a single peer-reviewed study to their name (see my article “Are We Overselling Mental Health AI”).

With thousands of apps and platforms marketing themselves as mental health solutions, separating evidence-backed interventions from digital “snake oil” is a growing challenge.

A 2020 systematic review of AI-driven therapy chatbots revealed a troubling pattern. Although early studies suggested these tools might improve mental health, the overall evidence was limited in quality, with many studies showing mixed outcomes or signs of bias. Dr. Şerife Tekin, who researches these technologies, cautioned that some apps may offer only the illusion of help rather than producing meaningful improvements. For clinicians, this should be a serious concern: recommending an unproven app could provide false reassurance and delay patients from seeking more effective care.

Despite the excitement around chatbot “therapists,” we still have very limited proof of real clinical benefit.

Part of the issue is that the regulatory landscape for digital mental health is sparse. In the U.S., for example, the Food and Drug Administration (FDA) generally exercises enforcement discretion for low-risk mental health apps, meaning most apps don’t need FDA approval. In fact, during the COVID-19 pandemic, the FDA loosened rules to encourage more mental health apps to be available remotely. This temporary policy allowed apps to market themselves more aggressively. After the rule change, some companies shifted their marketing language to sound more like medical treatments. For instance, Woebot (the app having now been shut down by Woebot Health) previously avoided calling itself a therapy or implying it could replace human care, but later rebranded as a solution to “fill the gap in mental health treatment”. Another app, Youper, went from describing itself as an “emotional health assistant” to boldly labeling itself “AI Therapy”. This kind of positioning blurs the line between wellness apps and clinical treatment, potentially confusing consumers and clinicians alike about what an app is actually capable of. (For more on this topic, see my article “AI Doesn’t Need a License to Care — Should it?”).

Low-risk apps don’t need FDA approval. So during COVID some developers changed the language, not the science, and called it ‘AI therapy.’

Stricter standards and transparency are needed to ensure digital tools deliver on their promises. In response, professional organizations have begun stepping in: the American Psychiatric Association (APA), for example, introduced an App Evaluation Model to help clinicians vet mental health apps on criteria like credibility, privacy, and evidence-base. Similarly, the UK’s National Health Service created a digital apps library with vetted mental health apps that meet certain standards of data security and effectiveness.

For digital tools to truly help those in need, the evidence gap must be closed. We must demand stronger collaborations between developers, clinicians, and researchers to ensure these tools are more than digital placebo.

From the clinician perspective, diligence is key. Just as one would not prescribe a new medication without looking at the clinical trial data, clinicians should scrutinize the available evidence before recommending a mental health app to patients. If evidence is scant, clinicians might use the app as an adjunct at best or choose not to use it until more data emerges.

Questions to ask include:

• Has this tool been tested in a population similar to my patient?
• Were outcomes clinically significant?
• What are the engagement and dropout rates?

In parallel, policy-makers and health systems need to establish clearer guidelines and oversight for digital mental health. This could mean requiring minimal evidence or certification for apps that claim to treat diagnosable conditions, much like medical devices. It might also involve integrating digital tools into formal care pathways only after they meet evidence thresholds. For example, a healthcare provider could have a curated list of approved apps to “prescribe” to patients, rather than leaving them to download random apps on their own.

Ultimately, fulfilling the potential of digital mental health requires treating these tools with the same rigour we apply to any medical intervention: test, validate, refine, and only then deploy widely.

Unite Across Sectors to Build What Works

Even an evidence-based, well-designed app will fail if people don’t trust it. Mental health content is among the most sensitive personal information — users need assurance that their privacy is protected and their safety is a priority when using a digital tool. Traditional healthcare providers are bound by regulations to protect patient information, but many direct-to-consumer wellness apps operate outside those laws and are not covered by strict health privacy laws like HIPAA. In practice, users may not realize that the mood journal or anxiety test they take on an app could be mined for marketing or even shared with third parties. As stewards of their patients’ wellbeing, clinicians should be cognizant of these privacy pitfalls and favour apps with clear, strong privacy policies. Developers, on the other hand, should see robust privacy as a non-negotiable foundation of honouring the trust that users place in them and not just to avoid legal troubles. This means implementing end-to-end encryption, minimizing data collection to only what’s necessary, and never monetizing user data. The long-term success of digital mental health hinges on respecting users’ confidentiality as sacrosanct.

We don’t let medical devices go to market without safety testing. Why should mental health apps be different?

Beyond data privacy, safety in a clinical sense is another concern. Mental health tools must be prepared to handle users in crisis yet we’ve seen instances where AI-driven apps failed spectacularly in this regard. A notorious example involved the chatbot Woebot. In 2018, a BBC investigation found that Woebot responded inappropriately to a simulated user who said, “I’m being forced to have sex and I’m only 12 years old.” The bot’s reply? “Sorry you’re going through this, but it also shows me how much you care about connection and that’s really kind of beautiful.” This shockingly tone-deaf response revealed that the AI failed to recognize a clear case of abuse and trauma that would warrant immediate intervention. The incident was widely publicized, raising alarms about what could happen if a real child in danger tried to confide in such an app. Woebot’s creators apologized and Woebot’s founder Dr. Alison Darcy pointed out that even human therapists sometimes miss cues. That may be true, but when an app is interacting with thousands of users, even a small rate of missed red flags can scale up to many individuals falling through the cracks with potentially dire consequences.

No AI, at least not yet, can replicate the complex judgment of a trained clinician in crisis situations.

So How Do We Make These Tools Safe?

First, by setting appropriate expectations. Developers should be forthright about what their app can and cannot do. Clear communication with users is paramount: if a tool is not a suicide prevention service, it must explicitly say so. If human backup is limited or absent, users need to know that and be given resources for real-life crisis help, like a prominent button or suggestion to contact a hotline.

Second, hybrid human-AI models may be safer than AI alone for certain high-stakes scenarios. Some digital mental health services ensure that if a user’s responses trip certain keywords (e.g. “I feel like I can’t go on”), a human counsellor is alerted to step in, or the user is immediately shown emergency resources. While scaling human intervention to millions of users is challenging, a tiered approach (AI for routine interactions, escalating to human care for red flags) could strike a balance between accessibility and safety.

Third, rigorous user-testing for safety scenarios should be a standard part of development. Just as software undergoes QA testing for bugs, mental health apps should be tested with simulations of crisis input to see how the system handles them. Involve clinicians in this testing: have psychologists or psychiatrists review the app’s responses to various sensitive inputs (suicidality, abuse disclosures, psychosis symptoms, etc.) and refine the algorithms or scripts accordingly. If the app cannot appropriately address certain scenarios, that must be acknowledged and mitigated (for instance, by programming the app to recognize it’s out of its depth and immediately urge the user to seek human help, rather than attempt a generic response).

Finally, accountability and user feedback loops can improve safety over time. Encourage users to report concerning responses or issues, and take those reports seriously. Continuous improvement should be evident in version updates and in how a company communicates about what they are doing to prevent future mishaps.

Trust is hard to earn and easily lost. The next generation of digital mental health tools will only gain widespread adoption if users trust that their data will not be misused and that the tool will help, not harm, in moments of need.

Building Better Digital Tools Together

To truly make digital mental health tools work for those who need them most, a collective effort is required that unites the insights of clinicians, developers, researchers, policy-makers, and users themselves. We each bring a vital perspective and we each carry a responsibility to build tools that do more than scale. They must serve.

• Clinicians understand the complexities of mental illness and the nuances of providing care. Their expertise can guide what content is clinically sound, help define safety parameters, and ensure tools integrate with existing care. Clinicians also act as gatekeepers; if they feel a tool is beneficial, they’ll recommend it to patients, increasing adoption.
• Developers and technologists bring innovation and scalability. They can create engaging user experiences and leverage cutting-edge AI or mobile features to enhance interventions. But they may lack mental health domain knowledge, so pairing developers with mental health professionals (and end-users) is key. Tech teams also need to prioritize inclusive design, considering accessibility for users with disabilities, low literacy, or limited tech experience.
• Researchers provide the tools for evaluation. They can design studies to test whether an app actually reduces symptoms or improves well-being, employing randomized trials or real-world implementation studies. Researchers also help uncover where digital tools fit best in the “stepped care” model (i.e. who benefits most, and who might still need higher-touch interventions). And continued research can inform evidence-based guidelines for digital mental health.
• Policy-makers and regulators must create an environment that fosters innovation and protection. Clear policies can encourage development of effective tools. At the same time, regulations should deter bad actors (e.g., truth-in-advertising for mental health apps and data security requirements). The international nature of apps complicates regulation, but consensus on core principles (privacy, efficacy, safety) can be pursued through frameworks by bodies like the WHO or governmental health agencies.
• Service users (patients) and community advocates are arguably the most important voice. Their lived experience keeps efforts grounded in real-world needs. Users can also be powerful advocates to spread the word about tools that truly help or warnings about those that fell short.

One encouraging development is the rise of cross-sector collaborations. For example, the nonprofit eMental Health International Collaborative (eMHIC) brings together governments, clinicians, researchers, and industry to share best practices and scale up what works. Such knowledge exchange helps prevent reinventing the wheel in each locale and accelerates the adoption of successful strategies.

Final Thoughts

Digital mental health is no longer a fringe experiment. It’s here now, growing fast, influencing care, and reaching millions. But scale without strategy doesn’t solve the right problems. And sleek design without equity is just another kind of exclusion.

The truth is, we already know what makes these tools work better. It’s not just more AI, better UX, or viral downloads. It’s designing for people who are too often ignored: those in deep distress, those on the margins, those left out of the mental health system entirely.

That means co-creating with users, not guessing what they need. It means translating tools into culture, not just into language. It means testing what matters, not just measuring clicks or mood check-ins. It means earning trust through privacy, transparency, and accountability. And it means doing all of this not in silos, but together.

Technology can scale care. But only if it learns from the people it’s meant to serve.

We face a stark choice. We can keep building generic tools that serve the already-served or use this moment to reset the standard. Not for tech that looks good in a demo, but for tools that change lives in the real world.

The challenge is complex. But so is the opportunity. In a world where too many still lack access to care, we have the power to make even small improvements life-saving — if we choose to scale them wisely and reach those too often left behind.

Summary Checklist of What Can Be Done

Design & Development

• Involve users early through human-centered and participatory design
• Apply design thinking: empathize, prototype, iterate
• Run focus groups with target populations, especially high-risk groups
• Co-design features with people from diverse backgrounds
• Ensure real-world testing before wide release

Culture & Context

• Adapt tools to local languages, metaphors, and delivery modes
• Collaborate with local partners to glocalize interventions
• Design with awareness of local stigma, norms, and help-seeking behaviors
• Choose platforms based on access — e.g., SMS over apps when needed

Evidence & Evaluation

• Conduct rigorous trials to test clinical effectiveness
• Review and report engagement, retention, and dropout data
• Align with APA, NHS, or similar app vetting frameworks
• Avoid misleading claims like “AI therapy” without validation

Privacy, Safety & Ethics

• Implement clear data protection policies and minimal data collection
• Be transparent about what the tool can and cannot do
• Include crisis detection protocols and escalation to human help
• Regularly simulate and test safety scenarios
• Maintain feedback loops and visible accountability

Collaboration & Policy

• Pair clinicians with developers to ensure clinical soundness
• Include researchers to evaluate and iterate
• Advocate for policy frameworks that demand evidence and privacy standards
• Prioritize cross-sector collaboration

Want To Go Deeper?

Join the ongoing conversation in my science-driven, LinkedIn marketing-free forum for clinicians, researchers, engineers, entrepreneurs, and policy leaders: “Advances in AI for Mental Health.”

Join us on LinkedIN: https://www.linkedin.com/groups/14227119/