At the beginning of the year, 30-year-old businesswoman Igihozo (who preferred to use only one name) was nursing a heartbreak and longed for a person she could talk to openly — one who wouldn’t judge her, but would listen, guide, and truly understand. Although she tried confiding in her best friend, it didn’t help; her friend couldn’t fully grasp what she was going through. Watching movies to distract herself didn’t help either, as it couldn’t ease the pain of losing a relationship she had invested in for years. ALSO READ: Mental health: Can AI-powered chatbots enhance access to care, combat stigma? In February, two weeks after the breakup, while working on a task, she sent her copy to ChatGPT to correct its grammar. After receiving the edited copy, she thought to herself: If AI can assist with my work, maybe it can help with my issues too. She started asking questions like: Why did he leave me? Was it my fault? Why do I still love him even after he hurt me? How can I stop thinking about him? Will this pain ever go away? Should I block him on social media? With the responses she received, Igihozo felt comforted without being judged and understood without needing to explain herself. People increasingly find AI to be a tool that offers quick comfort, responses, and help when dealing with mental struggles, an approach described as AI therapy. ALSO READ: Should you rely on ‘AI therapists’ for mental healthcare? Experts define AI therapy as mental health support delivered through artificial intelligence systems, such as chatbots or virtual assistants that use natural language processing and machine learning. These tools simulate therapeutic conversations, often drawing on Cognitive Behavioral Therapy (CBT) techniques to ask guiding questions, offer coping strategies, and prompt self-reflection. Dr. Celestin Mutuyimana, a psychotherapist at Baho Smile Institute, a psychotherapeutic and research center in Kimironko, says AI is useful but cautions against calling it therapy, especially in Rwanda’s context of psychological wounds. “I am involved in developing digital interventions, but these are different from what many call AI therapy. If you turn to AI expecting full therapy, it will tell you it’s only an intermediate solution and that you should look for more sustainable help elsewhere,” he said. Dr. Mutuyimana noted that unless digital solutions are developed by professionals who understand local realities and cultural experiences, they can only serve as temporary or emergency tools rather than lasting answers for people living with trauma. “AI is something common. It’s not really culturally sensitive to the problems of people in a given environment. Personal touch and human connection are important in mental health support. “When we use AI, it’s kind of a one-way conversation. For instance, I talk to AI, and it replies but there’s no emotional connection or human contact. In societies like Rwanda’s, which value collective belonging and sociability, this lack of connection can limit AI’s effectiveness.Self-control in collectivist communities requires a sense of belongingness, connectivity, and sociability, which is lacking in AI tools,” he explained. Dr. Mutuyimana also noted that AI itself often advises users to seek professional help, which signals its limits. He questioned how many people in Rwanda’s population of over 14 million could access AI tools, pointing to gaps in digital literacy and limited resources. By contrast, he noted, interventions designed by psychotherapists and adapted for mobile phones might reach far more people, since many already have smartphones connected to the internet that could be used for self-help support. “While AI therapy might help some, it cannot stand alone in terms of treatment, accessibility, or sustainability. It needs other sustainable solutions,” Dr. Mutuyimana added. Schadrack Niyibizi, a Software and AI Engineer based in Kigali, says AI could help address gaps created by the shortage of mental health professionals in Rwanda but it must be seen as a support tool, not a replacement for human care. “AI could support early detection of people suffering from mental health issues, for instance, by using models that ask different questions, it can identify signs that someone might be struggling. It could also offer recommendations, like suggesting they visit friends or seek professional help.” He added that technology could help mental health experts reach more people without meeting every individual face-to-face. “When there are only two or three mental health experts, AI can help them support many people by identifying those who may need help,” Niyibizi added. Niyibizi noted that mental health challenges vary for example, someone might be facing depression, and simply talking to another person, even if they’re not a trained expert can help them feel a bit better. AI could help identify such cases and suggest small supportive actions. However, AI in healthcare must be used carefully because we’re dealing with human life, and AI is just a tool, he added. According to Niyibizi, AI should be integrated into mental health services in Rwanda, countries like the United States and European countries, where they have many mental health professionals, still use AI to help detect people who are struggling. He explained that some people prefer to keep quiet or only talk to friends when they’re going through something but when interacting with a machine, they might feel more comfortable opening up. Philbert Murwanashyaka, the Head of Yali Labs, a Rwandan AI company in Kigali, described AI therapy as a new way for people to be treated, especially in the health sector. He noted that AI offers a type of personalised treatment plan, which people often look for when visiting doctors or therapists. As AI becomes more intelligent, Murwanashyaka said it is growing closer to real-life scenarios. He gave an example of Nvidia’s World Foundation Models that understand real-life physics, explaining that the more AI becomes intelligent, the more it can introduce new treatment options instead of people going directly to a therapist. Murwanashyaka added that there are things that require empathy and natural intelligence, which AI, as a machine, does not fully understand. He explained that large language models like OpenAI’s ChatGPT are trained on huge amounts of data and refined with advice from real consultants, making them capable of providing accurate information. Murwanashyaka also mentioned industry-specific tools like Lola AI, developed by Healthtracka, which offers menstrual health support via WhatsApp without needing to install an app. “One can register the number 0786880881 and interact directly,” he said. “Such innovations often come from under-resourced contexts, for example, in Africa, we don’t have the infrastructure, no GPU (Graphics Processing Unit) or TPU (Tensor Processing Unit) or expensive high-performance computing. Instead, creativity drives solutions, like smaller AI models that run directly on phones or USSD-based tools for feature phones without internet. I’ve seen a Tanzanian company do this,” he added. Murwanashyaka noted Cassava AI, a Zimbabwean company deploying GPUs in South Africa to help African governments access computing power, adding that some GPUs were deployed in June in data centers to help local AI efforts. “Companies in Rwanda are also exploring AI for health data analytics. I saw one project building a model for special disease tracking. This aligns with Rwanda’s new National Health Intelligence Hub, which collects data from various platforms to make health services more data-driven. For example, when more pneumonia cases appear in Gicumbi, AI can predict it, and authorities can mobilize the right drugs or support instead of waiting for the outbreak to spread.” “We can’t match the resources of developed countries, so we’re being innovative with what we have, like on-device AI as a growing area. Big companies like Apple, Google, and Meta are working on this too, to make models smaller and usable without cloud computing,” Murwanashyaka said. Dr. Jean Damascene Iyamuremye, the Director of Psychiatric Care at Rwanda Biomedical Centre (RBC), noted that since 2017, RBC has supported AI tools like RBC‑Mbaza, a chatbot offering COVID‑19 and health advice, including mental health tips in English and Kinyarwanda. “This open source AI system provides reliable health information and continually improves via machine learning,” he said. Comparing AI-based therapy to traditional therapy, Dr. Iyamuremye said the two are important and complementary, adding that one can’t work alone. He noted that AI-based therapy is accessible 24/7, helpful for those in remote areas or with busy schedules, since it offers uniform responses and can serve thousands simultaneously at low cost. Iyamuremye said that traditional therapy provides human connection, and trust, which influence outcomes, adding that therapists can adapt methods in real time to a client’s unique needs and non-verbal cues. Dr. Iyamuremye advised that traditional therapy is generally more effective for trauma, severe depression, personality disorders, and complex interpersonal issues. Clinical psychologist Christella Ishimwe from mHub Rwanda, a mental health clinic in Kicukiro, explained that while AI can help people identify symptoms and understand what they might be related to, it lacks human empathy. “When you open up to a person, they can read your body language, pick up on your emotions, and respond in a way that makes you feel truly seen and heard, something AI can’t do,” she said. Ishimwe noted that although AI tools can offer guidance, they often aren’t fully contextualized because they don’t have the capacity to feel or relate emotionally. “AI doesn’t replace the human ability to understand how someone feels and to offer genuine emotional support, which is essential,” she added.
