Imperial College London, the AI-powered mental health platform Wysa, and partners in India have launched a £5.3 million study funded by the Wellcome Trust to adapt and test an AI chatbot for adolescent girls struggling with anxiety and low mood in rural India, the university announced February 18, 2026. The project pairs Imperial’s health-data researchers with the Tata Institute of Social Sciences, the Milaan Foundation and the University of Cambridge to find out whether a tool already used by millions can be reshaped to reach some of the hardest-to-serve young people in the world’s most populous country.
Why Rural India, Why Girls
India is home to more than 253 million children and young people aged 10 to 19 — the largest adolescent population of any country on Earth — at a moment when, researchers note, roughly half of all lifetime mental health conditions first emerge before age 14, and suicide ranks among the leading causes of death for young people. Girls in rural areas face a particular bind: higher documented rates of anxiety and depression, compounded by restricted autonomy, limited access to phones and the internet, lower literacy, and family or community stigma that often keeps them from seeking help even when it exists nearby. India has only a small fraction of the mental health professionals that its population would require under World Health Organization benchmarks, and the shortage is most acute outside major cities, which is part of why researchers are looking at a scalable, low-cost digital tool rather than trying to train and deploy enough human counselors to match demand in the near term.
How Wysa Works
Wysa is not a new product — the platform already serves more than seven million users across 105 countries, including through partnerships with the UK’s National Health Service and Singapore’s Ministry of Health, and it combines an AI chatbot offering self-help exercises with escalation to human counselors when a user needs more than automated support. What’s new here is the attempt to culturally adapt that same clinically validated tool for a very different context: low literacy, limited connectivity, local languages and idioms, and family structures that often gatekeep a girl’s access to a phone in the first place.
Built on Evidence, But Untested at This Scale
The project leans on prior work: Imperial’s Professor Ceire Costelloe previously led an evaluation of Wysa inside NHS mental health services that helped establish a real-world evidence base for the tool. “Our role at Imperial is to ensure that AI-enabled interventions are properly evaluated using real-world data,” Costelloe said of the new project. Wysa’s chief clinical and R&D officer, Chaitali Sinha, who is a principal investigator on the study, said the goal is to “co-design a digital intervention that is not only clinically effective” but also acceptable to the girls it’s meant to serve. Tellingly, the project’s first phase isn’t deployment at all — it’s ethnographic groundwork mapping the cultural, social and technological barriers that shape whether girls can use a tool like this before anyone adapts the chatbot itself. That sequencing — study first, build second — is meant to avoid a common failure mode in global digital health projects, where a tool designed for one cultural and technological context is simply translated and dropped into another, only to see adoption collapse once it meets unfamiliar norms around privacy, family oversight or how emotional distress is discussed at all.
The Skeptics’ Case
Not everyone is convinced AI chatbots belong on the front line of adolescent mental health. The FDA’s own Digital Health Advisory Committee met as recently as November 2025 specifically to scrutinize generative AI mental health tools, with members pressing for reliable mechanisms to detect and escalate acute safety risks — the scenario where a chatbot, rather than a clinician, is the first thing a struggling teenager encounters. In a rural Indian setting, critics could point to a harder problem still: a chatbot requires a phone, data access and enough digital literacy to use it, meaning the AI tool risks reaching the relatively better-connected girls while bypassing the most isolated and vulnerable ones it was designed to help, and it cannot replace scarce human clinicians in a crisis. Researchers on the project acknowledge this tension directly, framing the entire first phase of the grant around identifying exactly which girls are likely to be excluded and why, rather than assuming a chatbot is automatically equitable just because it is cheaper to scale than a clinic.
What’s Next
Researchers plan to move from the current mapping phase into adapting Wysa’s content and interface, then piloting it in communities to measure whether it’s effective, acceptable and feasible — the standard bar for digital health tools before any talk of wider rollout. Wellcome’s director of mental health, Miranda Wolpert, framed the ambition as addressing anxiety and depression “in adolescent girls across rural India,” while Wysa chief executive Jo Aggarwal said the aim is for the tool to be “there for her not just in prevention, but in the moments when she’s truly struggling.” If the adaptation works, the model could be exported to other low-resource settings grappling with youth mental health gaps; if the access and trust problems prove too stubborn, it will be a cautionary tale about how far an AI chatbot can travel from the well-connected markets where it was built.
Photo: https://kaboompics.com/ / PEXELS via Pexels
