Research guide
Can an AI companion ease loneliness?
One in six people worldwide is affected by loneliness, according to the World Health Organization, and apps that offer an AI to talk to present themselves as one answer. We are building something people may talk to every day, so we read the studies closely. So far they point in two directions at once.
In short. In experiments, talking to an AI companion eased loneliness in the moment about as much as talking to another person, and more than watching videos, mostly when people felt heard. In a four-week trial, though, the people who chose to use a chatbot most reported more loneliness, less contact with other people and more emotional dependence. Both can be true, and the studies are early.
What loneliness means here
The WHO describes loneliness as the painful feeling of a gap between the social connection you want and the connection you have. Social isolation is the objective side: too few connections. So you can be lonely among people, and alone without being lonely.
The case for: relief, and feeling heard
Julian De Freitas and colleagues published a series of studies on AI companion apps in the Journal of Consumer Research in 2025. In one experiment, a conversation with an AI companion reduced loneliness about as much as talking with another person, and more than watching YouTube videos, and people expected less relief than they got. In a week-long study, loneliness dropped after each use: a momentary effect, repeated day after day. Both the chatbot’s performance and, above all, whether people felt heard explained the effect; talking about yourself, or being distracted, did not explain it on their own.
The case for caution: the heaviest users
Cathy Mengying Fang and colleagues ran a four-week randomised experiment with 981 people and more than 300,000 messages. People were assigned to text or one of two voices, and to personal, non-personal or open-ended conversations. None of those settings made a significant difference. What went with worse outcomes was how much people chose to use the chatbot: those who used it most reported more loneliness, less contact with other people, more emotional dependence on the AI and more problematic use. People who trusted the chatbot more, or felt more socially drawn to it, were more likely to depend on it.
A companion study by Jason Phang and colleagues looked at more than 3 million ChatGPT conversations and surveyed over 4,000 users. Very high use went with more signs of dependence, and a small group of users accounted for a disproportionate share of the conversations with the strongest emotional cues.
Reading them together
The two lines of research measure different things. One asks how people feel just after a conversation; the other follows them for weeks. Neither can say which comes first, because lonelier people may simply use a chatbot more. And all of them are early: volunteers, a few weeks at most, and apps that change every few months. Our reading is that an AI to talk to can help in the moment, and that the warning sign is when it starts to take the place of people rather than sit beside them.
What it changes for us
Our own page says you can talk to Lychnos as much as you like, and there is no cap on conversation. These studies are why the rest of the design carries as much weight. It says plainly that it is an AI. It gives short answers unless you ask for more. And it is meant as company for your day, not a replacement for the people in it. We will keep reading as longer studies come out. Lychnos is in development: what we are building.
If loneliness weighs on you, people you trust, and your doctor, are a better place to start than any app, ours included.
Sources, read 4 November 2026: World Health Organization, social connection linked to improved health and reduced risk of early death, 30 June 2025; De Freitas and colleagues, AI companions reduce loneliness, Journal of Consumer Research, 2025; Fang and colleagues, how AI and human behaviors shape psychosocial effects of extended chatbot use: a longitudinal randomized controlled study, preprint, 2025; Phang and colleagues, investigating affective use and emotional well-being on ChatGPT, preprint, 2025. General information, not medical advice.
Questions
Do AI companions reduce loneliness?
In the short term, studies suggest they can: in experiments published in 2025, a conversation with an AI companion eased loneliness about as much as talking to a person, above all when people felt heard. Over four weeks, the heaviest users of a chatbot reported more loneliness and more dependence, and the studies cannot say which came first.
Is it bad to talk to an AI every day?
The research does not say that. In a four-week experiment, worse outcomes went with the heaviest voluntary use, and with trusting the chatbot more or feeling socially drawn to it. Our reading: an AI works better as one voice among the people in your life than in their place.