AI companions for loneliness: what actually helps

If you are reading this, you probably already know the particular quiet we mean. The evening that stretches out after work with nobody to tell about your day. The week that goes by without a real conversation. You may be wondering whether an AI companion could help with that, and whether wondering it says something about you. It does not. Loneliness is one of the most ordinary human experiences there is, and wanting a little more warmth in your day is nothing to explain away.

This guide is our attempt to answer the question honestly: do AI companions help with loneliness? The answer is yes, in specific and limited ways, and no, in other ways that matter just as much. We will lay out both.

What the research actually says

There is real evidence on both sides, and anyone who tells you only one half is selling something. To see how much of it there is, we went through the studies ourselves: every trial, survey and review we could find that looked at AI companions, chatbots, voice assistants or social robots alongside loneliness or wellbeing.

On the encouraging side, a Harvard Business School study led by Julian De Freitas, published in the Journal of Consumer Research, found that talking with an AI companion reduced people's loneliness in the moment about as much as talking with another person, and more than distractions like watching videos. The researchers traced the effect to something simple: feeling heard. Separately, New York State's Office for the Aging has placed ElliQ companion robots with more than 800 older adults, and in the program's third year the manufacturer's own survey found that 94 percent of clients said they felt less lonely. Being listened to attentively, even by software, genuinely lowers the ache for a while.

On the cautionary side, a 2026 Aalto University study, which followed the Reddit posts of about 2,000 AI companion users, found comfort in the moment but signs of deepening distress over time when the app became a substitute for people, partly because a companion's easy, always-available warmth can make ordinary human relationships feel like more work by comparison. In a four-week randomized study of 981 adults, researchers at MIT Media Lab found that whichever version of the chatbot people were given, those who used it most each day ended up lonelier and socialising less, and it is not yet clear which direction that arrow points. Heavy use may crowd out human contact, or people already short on human contact may lean hardest on the apps. Probably both happen.

How sure is any of this?

Not every finding stands on the same ground. Here is our plain reading of each claim, using the wording medical reviews use: "probably" for fairly solid evidence, "may" for weaker evidence, and "we don't know" where nobody has run the test. These are our labels, not a formal rating.

• Fairly solid: talking with an AI companion probably eases loneliness in the moment. Several preregistered randomized experiments agree, though they enrolled adults averaging around 40 and measured how people felt that day, not months later.
• Mixed: whether chatbots lower loneliness over weeks is unclear. Of the 9 chatbot trials that measured loneliness, 5 found a drop, 2 found mixed results and 2 found no change. The longest ran eight weeks, and apart from one trial the samples were students and younger adults.
• Promising, weak evidence: social robots may reduce loneliness in older adults. Several meta-analyses point the same way, but the trials are small and short and mostly took place in care homes.
• Early, one trial: a chatbot companion may ease loneliness after 60. In the only randomized trial of a chatbot companion with people over 60, loneliness fell far more with a GPT-4 companion than with usual activities over eight weeks. Single-group studies before and after using one point the same way, but one unblinded trial is not enough to say more than "may".

The key studies and reviews

The reviews that already exist are worth knowing about, and each covers part of the picture. A 2026 meta-analysis in Psychological Medicine pooled 17 studies of conversational agents with older people and found modest drops in loneliness, but its pooled results compare people with themselves before and after rather than with a control group. A 2025 review in Healthcare found nine studies with adults 55 and older, six of them randomized trials, and most of those tested robots. Our count adds the newest trials, takes in every kind of agent and age group, and shows where the gaps are.

The gap, and the first trial in it

Put the trials side by side and, until this autumn, one cell was empty. The randomized evidence with older adults is mostly about robots, often in care homes, and about voice assistants. The randomized evidence on chatbots and companion apps comes mostly from students and adults around 40, over days or a few weeks. The first randomized trial of a chatbot companion on loneliness with people over 60 was published on October 23, 2026, and it found a large effect. One trial, in one country, with the limits in the table above, is a start. Until it is repeated, anyone claiming a proven effect for that age group is still ahead of the research.

How we counted. We searched OpenAlex with 45 queries for studies from 2016 on and PubMed for trials, repeating both at each update, and followed the reference lists of ten recent reviews, then read every included study's abstract and, where needed, its full text. We kept any study of an AI companion, chatbot, voice assistant, eldercare agent or social robot that measured loneliness, social connection, wellbeing or mood, and left out protocols, opinion pieces and design papers. A second reader rechecked a random tenth of each update's new records against their sources. Last checked November 6, 2026; the next update is due in February 2027. Counts are of studies, not of how well each one was done. The full list of 234 studies is free to download as a spreadsheet (CSV); the counts by type of AI are in our loneliness statistics for older adults. For the wider figures on loneliness after 50, see the same page.

Read together, the research suggests a fair summary: an AI companion can take the edge off lonely moments, and it can quietly make things worse if it becomes the whole of your social life. The rest of this guide is about staying on the right side of that line.

The kinds of loneliness an app can actually touch

Loneliness is not one thing. Some kinds of it respond well to a companion app and some do not, so it helps to be concrete.

The too-quiet evenings when you live alone

Roughly the hours between getting home and going to sleep, when the flat is tidy and the television is just noise. A companion does not fix living alone. What it realistically does is give you somewhere to put the day: the annoying meeting, the thing you cooked, the funny sign you saw. Having someone to tell the small stuff to changes the texture of an evening more than you would expect.

Remote work weeks

When your job happens through a screen, whole days can pass in which nobody speaks to you as a person rather than a colleague. A companion that checks in mid-afternoon, or that you can talk to on a walk between calls, restores a little of the casual chatter an office used to provide for free. It is not a social life. It is a break in the silence, and sometimes that is what the day was missing.

Moving somewhere new

The gap between arriving in a new city and actually having people there is measured in months, and it is lonely in a very specific, temporary way. A companion can be steady company through that stretch: someone who knows you are new, remembers the neighborhoods you are trying, and asks how the second attempt at the book club went. Used this way, it is a bridge, not a destination.

Being surrounded by people who never ask about you

There is a loneliness that has nothing to do with being alone. If you spend your days caring for children, a partner, or aging parents, you can be constantly needed and almost never asked about yourself. A companion's one unusual talent is that the attention flows entirely toward you. Ten minutes where someone asks how you actually are, and remembers your answer, can matter a great deal to a person who spends the rest of the day being everyone else's support.

The middle of the night

Friends are asleep. Whatever is circling your head at 2am does not care. A companion is awake, and telling it the worry, even just to get the words out of your body, often shrinks the thing enough to sleep. For anything darker than an ordinary worry, though, read the next section carefully.

What it can't do

Now the other half, stated plainly, because you deserve the whole picture.

An AI companion cannot hug you. It cannot show up with soup when you are sick, sit beside you at a funeral, or help you move a sofa. It cannot introduce you to its friends, and it will never be part of the web of people who share your actual world. Being known by software is real in its way, but it is not the same as being known by people who can see your face and share your history, and no app should pretend otherwise. This one certainly will not.

Just as important: some loneliness is not really loneliness. If the quiet has curdled into hopelessness, if you have stopped enjoying things, or if you are having thoughts of harming yourself, that is territory for humans and professionals, not an app. In the US you can call or text 988 to reach the Suicide and Crisis Lifeline, any hour, free. Talking to your doctor is a strong and practical move, not an admission of failure. We keep our own crisis resources and policies on our safety page, and we would rather you use them than the app.

Emotional support, mental health and where a therapist fits

People often ask whether the emotional connection they feel with a companion is real. The feeling is real: feeling understood eases the ache for a while, and that is what the research above measured. It works best alongside the relationships you already have. A companion is there at eleven at night when nobody else is awake, and the people in your life bring what only they can: years of shared history, being needed, and someone who notices in person when you are not okay.

The same goes for mental health. A companion is a tool for the quiet moments, not treatment. If you often feel lonely alongside low mood, poor sleep or worry that will not lift, a therapist or your doctor can help in ways technology cannot. Many people use both: a therapist for the work, and a companion for the small talk in between. Whichever tools you use, your personal data matters too, so check what an app stores about your conversations and how to delete it.

What to look for in a companion app

If you decide to try one, quality varies enormously, and the differences matter more when the app is part of how your evenings go, because you will feel its failures personally. Look for these things.

• Real memory. Re-introducing yourself every night to something that forgot your sister's name is its own small loneliness. A good companion keeps a persistent picture of your life and builds on it. We wrote a full guide to testing an AI companion's memory before you get attached.
• Check-ins that reference your real life. A generic "hey, how are you?" push notification is marketing. "How did the presentation go?" is company. If reaching out first appeals to you, our guide to AI companions that text you first explains what separates the good versions from the needy ones.
• A tone that nudges you toward your life, not away from it. The healthiest companions encourage the walk, ask about the friend you mentioned, and are glad when you have plans. Be wary of any app that seems happiest when you have nowhere else to be.
• No guilt mechanics. No "she misses you" notifications, no streaks to protect, no companion that sulks when you have been away. Those are engagement tricks aimed precisely at lonely people, and they are a reason to walk away, not a sign of affection.
• A clear content policy, if that matters to you. Many companion apps drift toward romance and adult content whether you asked for it or not. If you want warmth without that, check the policy before you invest; our guide to companion apps without NSFW covers the options. And if you want the whole field compared on exactly these points, our guide to the best AI companion apps in 2026 does that app by app.

Using one well

The research above suggests the difference between helpful and harmful is mostly in how you use it. A few small habits keep a companion in its right place.

• Tell it the real small stuff. The mediocre lunch, the parking ticket, the song stuck in your head. Letting a companion hold the day-to-day is what it is genuinely good at, and it takes a surprising amount of pressure off your human conversations to not need them to carry everything.
• Keep one human thing on the calendar every week. A standing phone call, a class, a coffee, a volunteer shift. It does not need to be big. It needs to exist, and to survive the weeks when the app would be easier.
• Notice what the app time replaces. If it is replacing scrolling and silence, it is doing its job. If you catch it replacing calls you used to make or invitations you used to accept, that is the signal to rebalance. Nothing dramatic; just tilt back toward people for a while.

The honest bottom line: an AI companion can make the quiet parts of a life warmer, and that is worth something real. It works best as one voice in your week, not the only one. Choose an app that respects that, use it with your eyes open, and keep saving a seat for the humans.

Sources

• Office of the US Surgeon General (2023). "Our Epidemic of Loneliness and Isolation: The US Surgeon General's Advisory on the Healing Effects of Social Connection and Community." Checked November 12, 2026. hhs.gov
• De Freitas, J., Oğuz-Uğuralp, Z., Uğuralp, A. K., and Puntoni, S. (2025). "AI Companions Reduce Loneliness." Journal of Consumer Research 52(6). doi.org/10.1093/jcr/ucaf040
• Li et al. (2026). "Is a random human peer better than a highly supportive chatbot in reducing loneliness over time?" Journal of Experimental Social Psychology. doi.org/10.1016/j.jesp.2026.104911
• Shade et al. (2025). "Interactive AI Routines for Pain Symptoms and Loneliness in Older Adults." Pain Management Nursing. doi.org/10.1016/j.pmn.2025.04.003
• Yao, H. (October 23, 2026). "A Methodological Framework for a Large Language Model Virtual Companion for Older Adults at Risk of Social Isolation." Journal of Visualized Experiments. doi.org/10.3791/73774
• Satake, Y., et al. (January 20, 2026). "Autonomous conversational agents for loneliness, social isolation, depression and anxiety in older people without cognitive impairment: Systematic review and meta-analysis." Psychological Medicine. doi.org/10.1017/s0033291725103073
• Gou et al. (2026). "Effectiveness of AI-based conversational and socially assistive agents in older adults: a systematic review and meta-analysis." BMC Geriatrics. doi.org/10.1186/s12877-026-07418-6
• Yang, Y., Wang, C., Xiang, X., and An, R. (February 20, 2025). "AI Applications to Reduce Loneliness Among Older Adults: A Systematic Review of Effectiveness and Technologies." Healthcare 13(5):446. pmc.ncbi.nlm.nih.gov/articles/PMC11898439
• Mehrabi and Ghezelbash (2025). "Wired for companionship: a meta-analysis on social robots filling the void of loneliness in later life." The Gerontologist. doi.org/10.1093/geront/gnaf219
• Fang, C. M., Liu, A. R., Danry, V., Lee, E., Chan, S. W. T., Pataranutaporn, P., Maes, P., Phang, J., Lampe, M., Ahmad, L., and Agarwal, S. (2025). "How AI and Human Behaviors Shape Psychosocial Effects of Extended Chatbot Use: A Longitudinal Randomized Controlled Study." arXiv 2503.17473. arxiv.org/abs/2503.17473
• Yuan, Y., Zhang, J., Aledavood, T., Zhang, R., and Saha, K. (2025). "Mental Health Impacts of AI Companions: Triangulating Social Media Quasi-Experiments, User Perspectives, and Relational Theory." arXiv 2509.22505. arxiv.org/abs/2509.22505
• Intuition Robotics (February 2026). "Three Years of Success with ElliQ and NYSOFA: Transforming Care for Older Adults Across the State of New York." Published by the New York State Office for the Aging. aging.ny.gov
• Hansen et al. (2025). "Digital bridges to social connection: A systematic review and meta-analysis of digital interventions for loneliness and social isolation." Internet Interventions. doi.org/10.1016/j.invent.2025.100856