An AI bot developed in South Korea calls tens of thousands of seniors every day, asks about their well-being, reminds them to take medicines, and analyses their answers. Could a few minutes of conversation help early identify health and social problems before they become emergencies?
Social screening on a mass scale
South Korea is one of the world's fastest-aging countries. More than one in five residents is already over 65, the number of seniors living alone is growing, and the health and social care system faces shortages of doctors, nurses and social workers. Many older people live alone, which is associated with a higher risk of death. When no family members call to check if everything is fine, AI can.
Developed by Naver Cloud, South Korea's generative AI system Talking Buddy calls seniors and conducts conversations lasting a few minutes. It asks about how they feel, whether they have taken their medicines, how they slept, whether they have been physically active, and whether they are experiencing any problems. The system analyses the conversation and alerts a human social worker when it detects something unusual.
The system has already been involved in cases in which an older person's health was at risk. In 2024, 77-year-old Chung Yun-hee received a call while suffering severe abdominal pain and vomiting. She managed to say only a few words before ending the conversation. The AI nevertheless alerted a social worker, who arranged help. Chung was taken to hospital and underwent emergency surgery for an acute hernia.
For some seniors, however, the value of the call goes far beyond medical support. Chung said she rarely had an opportunity to talk with her family. The AI listened and responded, giving her a feeling that somebody was paying attention to her.
Some seniors talk to the system about their problems. Others play the piano for it or invite it to lunch, even though they know that the caller is a bot.
Let’s be honest: Loneliness is a health problem
The technology is being developed against the background of a much larger problem: loneliness among older people.
A large 2025 systematic review and meta-analysis looked at 86 prospective and longitudinal studies involving older adults. It found that loneliness was associated with a 14% higher risk of death from any cause. Social isolation was associated with a 35% higher risk, while living alone was associated with a 21% higher risk.
Studies from Europe show how widespread the problem is. A 2024 meta-analysis covering 70 studies and more than 462,000 older adults estimated that around 23% of older people in Europe experience loneliness.
For example, in 2025, the Netherlands had 3.76 million residents aged 65 or older, accounting for 20.8% of the population. Almost 932,000 were aged 80 or above. Living alone does not automatically mean being lonely. But when there is no regular contact with family, friends or care workers, a relatively simple phone call can become an important point of contact.
Human-led vs. AI-led companionship for seniors in the Netherlands
Let’s say a similar, human-led social screening is introduced in the Netherlands to fight loneliness and detect potential health problems early. Is it realistic to build such a service? If every Dutch resident aged 65 or older received a five-minute call every day, the system would have to make about 3.76 million calls a day. That is around 313,000 hours of conversation every day, or about 114 million hours a year.
If people made the same calls instead of AI, the scale would be very different. Even assuming that a full-time worker could spend five hours a day actually talking to seniors, the service would require roughly 63,000 full-time workers. At eight hours of conversation per working day, the figure would still be around 39,000. In reality, the number would be higher once holidays, sick leave, administration and preparation were included. The average basic salary of a social worker in the Netherlands is currently about €4,216 per month, according to salary data updated in August 2026. That would put the gross wage bill for 63,000 full-time workers at roughly €3.2 billion a year, before employer contributions and other employment costs.
€3.2 billion is about 2.7% of the Netherlands' total annual healthcare spending and almost 9% of what the country spends on long-term care. This is the basic economic argument for AI calling. A human social worker cannot make thousands of calls at once. An AI system can conduct many conversations in parallel and pass only cases that require human attention.
Europe is also testing AI calls, but the evidence is still limited
South Korea is not the only country experimenting with AI phone calls for older people. Madrid City Council has developed PALOMA, an AI system that calls people over 75 who live alone to identify unwanted loneliness and social needs. It conducts a short conversation and uses the answers to assess the person's situation. Cases that require attention are reviewed by municipal professionals.
The pilot made more than 5,000 calls and completed more than 2,000 conversations. According to the OECD, 646 people requested follow-up from municipal services, while the system identified cases requiring further intervention.
Finland is testing a similar idea. Aapo, developed by Vivonep, makes regular phone calls to older people and talks with them about everyday life, memories and interests. The service works through an ordinary telephone, so the user does not need a smartphone or an app. It is designed as additional support rather than a replacement for family members, carers or other human contact.
The South Korean system also shows the technology's limits. Generative AI can misunderstand what a person says, interrupt a conversation or make promises it cannot keep. It cannot decide on its own what kind of medical or social intervention a senior needs.
Another question is whether people in other countries would be as willing to talk to an AI. Acceptance of the technology varies by country and culture. South Korea is already ahead of the EU in business use of AI. According to the OECD, 30% of Korean enterprises used AI in 2023, compared with an EU average of 13.5% in 2024.
Evidence that AI conversations actually improve people's lives remains limited. A 2026 systematic review and meta-analysis published in BMC Geriatrics examined eight randomized controlled trials involving 611 older adults. Three trials, involving 190 people, measured loneliness. The analysis found no statistically significant reduction in loneliness compared with control groups. The researchers rated the evidence for loneliness as low certainty and noted substantial differences between the studies.
AI calls may provide companionship, identify problems, and make it easier for human care workers to reach people who need help. But the evidence so far does not support presenting a few minutes of conversation with an AI as a solution to loneliness.
References
BMC Geriatrics (research)
