Watch: our founder on who actually uses AI companions, why they keep quiet about it, and what the loneliness research found. Written version and sources.
The short version. Yes, in specific and limited ways. In a Harvard Business School study, talking with an AI companion eased loneliness in the moment about as much as talking with another person, and the effect came from feeling heard. The same research field warns that heavy use as a substitute for people can make things worse over time.
A companion app can take the edge off too-quiet evenings, weeks without a real conversation, and the middle of the night. It cannot carry grief or a crisis, and it should not replace the calls and invitations you already have. Look for memory that holds across weeks, check-ins that mention what you actually said, and no guilt mechanics.
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.
The research in brief. We found 179 studies, 102 of them published in 2025 or 2026. Only 18 were randomized trials that measured loneliness: 8 tested social robots, 7 tested chatbots or companion apps, and 3 tested voice assistants or similar agents.
Ten of those 18 trials enrolled people over 60. All ten tested robots or voice assistants. None tested a chatbot or companion app with people over 60, and none of the 33 companion-app studies we found had an older-adult sample at all.
Studies checked up to October 10, 2026. Details of how we counted are at the end of this section.
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 reported that 94 percent of clients said they felt less lonely. These are not small findings. Being listened to attentively, even by software, genuinely lowers the ache for a while.
On the cautionary side, a 2026 Aalto University study, covered by Forbes, described AI friends as a band-aid: comforting in the moment, but capable of deepening distress over time if they become a substitute for people, partly because a companion's easy, always-available warmth can make ordinary human relationships feel like more work by comparison. Other researchers have found that the heaviest users of AI companions tend to report feeling less supported by their friends and family, 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 7 chatbot trials that measured loneliness, 3 found a drop, 2 found mixed results and 2 found no change. The longest ran eight weeks, and the samples were mostly 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.
- Not tested yet: what a chatbot or companion app does for loneliness after 60. Single-group studies before and after using one suggest a modest drop. No randomized trial in our search has tested one with people over 60.
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 every one of its studies compared people with themselves before and after, with no 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.
| Study | Design and who took part | What it found |
|---|---|---|
| De Freitas et al., Journal of Consumer Research, 2025 | Randomized experiments with an AI companion built for the study; US adults averaging about 40, 1,072 in the week-long study | 15 minutes with the companion eased loneliness about as much as chatting with a person; daily use for a week gave repeated short-term drops |
| Li et al., Journal of Experimental Social Psychology, 2026 | Randomized trial, 2 weeks; 296 first-year students texted a supportive chatbot, a random fellow student, or wrote a journal | Texting a fellow student reduced loneliness; the chatbot lifted mood in the moment but did no better than journaling |
| Shade et al., Pain Management Nursing, 2025 | Randomized trial, 12 weeks, two versions of voice-assistant daily routines; 50 adults 60 and older living alone with chronic pain | Less depression in both groups, and significantly less loneliness with the fuller routine |
| Satake et al., Psychological Medicine, 2026 | Meta-analysis of 17 before-and-after studies of conversational agents (9 robots, 8 non-robot) with older people | Mild to moderate drops in loneliness and depression, and no study reported a worsening; no control groups |
| Gou et al., BMC Geriatrics, 2026 | Meta-analysis of 8 randomized trials of AI conversational and socially assistive agents; 611 older adults | A small drop in depressive symptoms; no significant effect on loneliness |
| Yang et al., Healthcare, 2025 | Systematic review of 9 studies with adults 55 and older (6 randomized); 6 social robots, 2 voice assistants, 1 digital human | 6 of 9 found significant reductions in loneliness, mostly with robots |
| Mehrabi and Ghezelbash, The Gerontologist, 2025 | Meta-analysis of 19 studies of social robots; 1,083 older adults | Significantly less loneliness, with bigger effects in care homes than among people living independently |
| Hansen et al., Internet Interventions, 2025 | Meta-analysis of 40 randomized trials of digital programs for loneliness; 6,062 people of all ages | Robotic pets reduced loneliness; the authors named AI social chatbots as the main gap in the research |
The gap nobody has filled
Put the trials side by side and one cell is empty. The randomized evidence with people over 60 is about robots, mostly in care homes, and about voice assistants. The randomized evidence on chatbots and companion apps comes from students and adults around 40, over days or a few weeks. Nobody in our search has yet run a randomized trial of a chatbot or companion app on loneliness with people over 60. Until someone does, anyone claiming a proven effect for that age group is ahead of the research.
How we counted. We searched OpenAlex with 45 queries for studies from 2016 on, ran PubMed searches for trials, and followed the references of four 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 pass rechecked a random tenth of the 179 records against their sources. Last checked October 10, 2026. Counts are of studies, not of how well each one was done. For the wider figures on loneliness after 50, see our loneliness statistics for older adults.
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.
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.
How My Softly approaches this
My Softly is our app: an AI companion app for iPhone that remembers what you tell it and asks about your day, for the days nobody else does. We built it for the evenings this guide is about.
We did not build My Softly as a cure for loneliness, because we do not believe an app can be one. We built it to be someone to share the day with: a companion with real long-term memory, so the small things you mention actually stay known, and gentle check-ins that can reach out first about your real life. It keeps quiet hours, and if you go quiet, it steps back rather than chasing you. There are no guilt mechanics of any kind, no streaks, no "I miss you" notifications engineered to pull you back. It is for adults 18 and over, it never starts sexual talk on its own, and it is available now on the App Store for iPhone, with free messages to start so you can judge it for yourself. More in our FAQ and on the home page.
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
- 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
- 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
- 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
Frequently asked questions
Do AI companions actually help with loneliness?
In specific, limited ways, yes. A Harvard Business School study found that talking with an AI companion reduced loneliness in the moment about as much as talking with another person. The same research field warns that heavy use as a substitute for people can make things worse. The difference is in how you use it.
Has an AI companion been tested in a clinical trial with older adults?
Not yet, as far as we can find. Of 179 studies we checked up to October 10, 2026, 18 were randomized trials that measured loneliness, and 10 of those enrolled people over 60. All ten tested social robots or voice assistants. The chatbot and companion-app trials enrolled students and adults around 40, and the studies of chatbots with older adults so far compare people with themselves before and after, which shows a modest drop in loneliness but cannot prove the app caused it.
Is it unhealthy to use an AI companion for loneliness?
Not in itself. It becomes a problem when the app replaces calls you used to make or invitations you used to accept. Keep one human thing on the calendar every week, notice what the app time replaces, and treat the companion as one voice in your week rather than the only one.
What kind of loneliness can an AI companion help with?
The everyday kinds: too-quiet evenings when you live alone, remote-work weeks without a real conversation, the months after moving somewhere new, being surrounded by people who never ask about you, and the middle of the night when friends are asleep. It cannot help with grief-level or crisis-level loneliness, which needs people and professionals.
What should I look for in a companion app if I'm lonely?
Real memory that carries your life across weeks, check-ins that reference what you actually said, a tone that nudges you toward your life rather than away from it, no guilt mechanics such as "she misses you" notifications or streaks, and a clear content policy if you want warmth without adult content.
Can an AI companion replace human friends?
No, and any app that suggests it can is not being straight with you. A companion cannot sit beside you at a funeral or introduce you to its friends. It can take the edge off lonely moments and give the small stuff of your day somewhere to land, which is worth something real.
What if my loneliness feels like more than loneliness?
If the quiet has turned into hopelessness, you have stopped enjoying things, or you are having thoughts of harming yourself, that is territory for humans and professionals, not an app. In the US, call or text 988 to reach the Suicide and Crisis Lifeline at any hour. Our safety page lists more resources.