Turkish-language Large Language Model-supported Voice Companion for Cognitive Monitoring via Social Artificial Intelligence in Older People Living Alone: A Single-Arm Pilot Feasibility Study

As the world's population ages, a quiet but growing challenge is reshaping how we think about health and community. Millions of older adults now live alone, often facing a double burden: the deepening isolation that comes from shrinking social circles, and the slow, sometimes invisible changes in memory and thinking that can accompany advanced age. Loneliness is not merely a feeling of sadness; it is a state linked to serious physical and mental health risks, including depression and heart disease. At the same time, detecting early signs of cognitive decline is difficult because standard medical tests require a doctor's visit, which can be hard to schedule or attend for those living in remote areas or with limited mobility. For decades, researchers have wondered if technology could bridge this gap, offering both a friendly voice to combat isolation and a gentle way to keep an eye on mental sharpness, all from the comfort of a home.

In a recent pilot study, a team of researchers in Turkey tested a new approach to this problem using a mobile application called iShe. This was not a simple chatbot that follows a rigid script, but a voice companion powered by a large language model, a type of advanced artificial intelligence capable of understanding and generating human-like conversation. The system was designed specifically for Turkish speakers, allowing it to engage in natural, empathetic dialogue while quietly weaving in questions to check a user's cognitive status. The researchers wanted to see if older adults living alone could use this technology daily without frustration, and whether the experience might improve their feelings of loneliness and depression while providing reliable data on their thinking skills.

The study involved ten volunteers, all aged 65 or older, who lived by themselves and had no prior diagnosis of dementia or significant memory loss. Over the course of one month, each participant was asked to have a daily conversation with the app. The interface was built to be simple: a large button to start and stop the recording, with the device speaking loudly through its own speaker so the user did not need to wear headphones or type anything. During these sessions, the app would chat about daily life, remember previous topics to make the conversation feel personal, and then gently ask ten specific questions about orientation, memory, and calculation. These questions were based on a standard medical test known as the Short Portable Mental Status Questionnaire, but they were delivered as part of a natural chat rather than a formal exam.

The results of this small trial were striking in their simplicity and consistency. Every single participant finished the full month of the study, completing all 300 scheduled sessions. On average, each daily conversation lasted nearly 19 minutes, a duration that suggests the users were genuinely engaged rather than just going through the motions. Only two people needed help from the research team to fix a technical issue, and that was during the first week; the rest managed the technology entirely on their own. This high level of participation indicates that the design successfully removed the barriers that often stop older adults from using complex digital tools.

Beyond the technical success, the study observed meaningful changes in how the participants felt. At the start of the month, the group reported moderate levels of loneliness and some symptoms of depression. By the end of the 30 days, their scores on standardized scales for these conditions had dropped significantly. The average loneliness score fell from 8.5 to 5.7, and the average depression score dropped from 4.5 to 2.3. While the study was too small to prove that the app caused these improvements, the direction of the change aligns with the idea that having a consistent, attentive listener can ease the burden of isolation. The participants also showed no signs of cognitive decline during the month; their scores on the in-app memory questions remained stable and matched the results of the face-to-face tests given by doctors at the beginning and end of the study.

The researchers were careful to note that this was a first step, a feasibility study designed to see if the idea worked in the real world rather than to declare a final cure. The group of participants was small and mostly consisted of women with university degrees, which means the results might look different for a broader, more diverse population. Furthermore, because there was no control group of people who did not use the app, it is impossible to say for certain that the improvements came solely from the technology and not from other factors, such as the novelty of using a new device or the attention received from the research team. The study also acknowledged that the artificial intelligence system was not yet a fully regulated medical device, and it lacked some of the safety filters that would be required for widespread clinical use.

Despite these limitations, the study offers a promising glimpse into the future of elder care. It demonstrated that an artificial intelligence voice companion can be used daily by older adults living alone without causing confusion or frustration. It showed that such a system can simultaneously provide social connection and gather reliable data on cognitive health, two goals that are often pursued separately. The fact that the participants' memory scores on the app matched the scores given by human doctors suggests that digital tools could one day serve as a continuous, low-burden monitoring system, catching early signs of trouble long before a crisis occurs. As the population continues to age, innovations like iShe suggest that technology may not just be a tool for efficiency, but a genuine partner in maintaining dignity, connection, and health for those who live alone.