Karin Andrea Stephan has a simple explanation for why she spent years building an AI-driven mental health companion.
Growing up she had always wished to have something to help her through difficult moments, she says.
“I didn't have that,” says Stephan, co-founder and chief operating officer of Earkick. “But I realised how valuable it would have been.”
She is not the only person in the industry whose motivation is personal. Across the platforms grouped under the label “AI mental health”, founders and executives describe being drawn to the sector by what they witnessed around them.
They say a gap existed between what psychology has discovered about well-being and the kind of help ordinary people could access.
Sidekick that did not exist
Earkick, founded in 2021, describes itself as an AI mental health companion built for responsibility and trust. Users can collaborate through voice, video or text. The app records interactions, correlates it with data such as sleep and movement, and responds through a panda avatar that remembers past conversations.
Stephan is careful about what she calls it: not a therapist, not an assistant, not a companion in the way the word is usually understood. “It is not built for entertainment, it is not there to show ads or make you addicted,” she says. “It is built to help you navigate moments in life that feel complicated, confusing or frustrating. At the same time, it is always trying to reconnect you with real life and your sense of purpose.”
One of the platform's design choices is that it does not aim to keep users engaged for longer than they want. A 2025 Harvard Business School working paper analysed 1,200 farewell responses across six of the most downloaded AI companion apps and found that five out of six used emotionally manipulative tactics when users tried to end a session. Earkick, Stephan says, is designed to work differently.
When a user shows signs of crisis, the response is designed to go beyond a generic redirect. Rather than disengaging, the app continues to support the user while encouraging them to speak to a human, whether a therapist or someone they trust.
Problem that could not wait
Khawla Hammad arrived in the sector through a combination of personal observation and data. Before founding Takalam, she worked in government roles across marketing, communications and strategy in the UAE. She had always wanted to build something meaningful. What she found when she started looking further into the subject was a region carrying a significant mental health burden with few accessible ways to address it.
“I saw family members and friends go through challenging situations,” she says. “When I looked into the data, I realised that more than 100 million people in our region were already experiencing depression or anxiety. And that was before Covid. After the pandemic, the need only grew.”
At the time, she says, the solutions available offered little in the way of convenience, privacy or variety. Those three gaps became the foundation of Takalam, now in its fifth year and serving users across more than 50 countries. The platform combines access to more than 80 licensed psychologists from varied cultural and linguistic backgrounds with AI-powered tools, including its companion Aila, trained by clinicians and operating within UAE data privacy regulations.
“From day one, we recognised that mental health is a human-driven field and it must remain that way,” she says.
Aila's role is to support users between therapy sessions, guide them through exercises, assist with administrative tasks, and help match them with the right therapist. It is not designed to replace clinical care. The platform also uses AI for screening and analysis, gathering more comprehensive data over time to help clinicians make better-informed decisions. “All insights are reviewed and validated by professionals before being shared with users,” Hammad says.
On the question of dependency, she draws a direct parallel with therapy itself. “Even with human therapy, more than one hour per day is not considered healthy. We apply similar principles. Aila usage is limited to ensure it remains a supportive tool rather than a substitute for therapy.”
Helping science reach people
Dr Xuan Zhao arrived at this problem from a different direction. She had spent years as a social psychologist working in university settings before co-founding Flourish Science in 2022 with Dr Julie Kasha.
The founding question, she says, was straightforward: why does so much validated psychological knowledge never reach the people who need it?
“There is so much knowledge about what we can do to improve our daily lives, to regulate our emotions and manage stress that is backed by scientific research but is not reaching the public,” she says.
“We wanted to figure out how to create an accessible, interactive format where people can find the most relevant knowledge and figure out what actually works for them.”
The result is Sunny, Flourish's AI well-being coach. The platform draws on motivational interviewing, cognitive behavioural therapy, positive psychology, mindfulness and neuroscience. But the approach is not prescriptive. Rather than handing users a set of exercises, Sunny asks questions.
“Think of it like a doula,” Dr Zhao says. “The kid is already in there. The doula's work is to help facilitate the birth. Motivational interviewing is the same idea.”
The platform recently underwent a third-party safety evaluation using VERA-MH, an open-source automated mental health AI safety benchmark developed through collaboration among clinicians, suicide-prevention experts, ethicists and AI developers. Flourish said the audit found it scored 21 points higher than ChatGPT and 49 points higher than Gemini on a 100-point clinical safety scale.
The crisis protocol goes further than directing a distressed user to a helpline number. The app distinguishes between passive suicidal ideation and active planning. It checks whether the person has a safety plan in place.
“Those are specialised workflows I have not seen from other mental health AI tools, or from general purpose AI,” Dr Zhao says.
The platform is designed, she explains, not to create a bond between a person and an AI but to help users connect better with themselves and with the people already in their lives.
“We don't need AI friends,” she says. “We need more human friends.”
Early voice in the room
Wysa was founded in Bengaluru around 2015, at a time when neither AI nor mental health were widely discussed, as Pranav Gupta puts it,
That makes it one of the earliest platforms of this type, according to Gupta, who is Wysa's global head of commercial and partnerships.
The platform was built around the idea that mental health care was too reactive. People sought help only when a problem became severe. Wysa wanted to change that by focusing on everyday well-being, using cognitive behavioural therapy techniques delivered through a chatbot in short sessions. No personal data is collected during onboarding. Users enter a nickname and begin.
“Think of Wysa as a mental health gym,” Gupta says. “It helps people build resilience over time.”
Since Covid, the platform has moved beyond its consumer app into partnerships with employers, insurers and public health systems, including the NHS in the UK and public health authorities in Singapore. More than 40 peer-reviewed studies and several randomised controlled trials back its effectiveness, showing reduced symptoms of anxiety and depression. The platform has also been recognised by the Mozilla Foundation as among the most privacy-focused apps in this space.
On safety, Wysa uses what Gupta calls a neuro-symbolic approach: rule-based clinical safeguards sitting alongside language models.
When optimism is not enough
Dr Prince Kouamelan Kouassi, chair of the department of psychology and assistant professor at the American University in Dubai, has concerns about the pace of adoption of the technology and what it might mean for a region still building its mental health infrastructure. He is not opposed to the technology, but what concerns him is the sequencing.
The UAE's mental health system is improving, with more psychologists, more clinics and more resources, he says. But the number of people seeking help is rising just as fast. Introducing a new layer of technology into a system that is still finding its footing risks making things harder, Dr Kouassi warns.
“If we introduce AI into a system that is already not fully optimised, we risk complicating the situation,” he says.
His second concern is about the fit. The models that underpin most mental health tools were developed in western settings and carried over to this region without being adapted to local ways of communicating, disclosing and seeking help.
Stigma here, Dr Kouassi says, is not simply a matter of tradition or resistance. Research connects it directly to a lack of culturally responsive practice.
“There are existing cultural frameworks for dealing with mental health in the region,” he says. “These should be integrated with imported models to create something more locally relevant.”
Then there is the speed problem. People on the ground adopt new tools quickly, but research moves slowly. Policymakers wait for the research. In the gap between those two timelines things can go wrong, Dr Kouassi says.
On the question of whether AI could replace therapists, he says that is not the right way to frame the discussion.
“The goal should not be replacement but support. We need to focus on building effective mental health systems first. Only then can AI be introduced in a meaningful way,” he says.
Nour Ibrahim is a 2025-26 Rosalynn Carter Mental Health Journalism Fellow
