October 18, 2026 · 7 min read
AI Companions for Older Adults: What the Research Shows in 2026
A 2026 meta-analysis of 17 studies found mild to moderate loneliness benefits from AI agents for older adults. What the evidence shows, and its limits.
Most headlines about AI companions picture someone young, scrolling late at night. But some of the most careful research on talking machines has been done with people in their 60s, 70s and 80s, often living alone, often going whole days without a conversation.
So what does the evidence actually say about AI companions for older adults? Does talking to a robot, a voice assistant or a chatbot make any difference? Here is a clear, honest look at the 2026 research, what it can and cannot tell us, and how to think about it for yourself or someone you love.
The short answer: early research is encouraging but not conclusive. A 2026 meta-analysis in Psychological Medicine pooled 17 studies of robots, voice assistants and chatbots used by older people without cognitive impairment and found mild to moderate improvements in loneliness and depressive symptoms, with no study reporting that people got worse. But most studies were small, had no proper comparison group and were not blinded, so the authors call for more rigorous research. AI companions look like a useful extra source of conversation for some older adults, not a replacement for people, care or services.
Why this matters
Going without conversation is more common than many people realise. In October 2026, Age NI and Lidl Northern Ireland reported that a quarter of older people in Northern Ireland sometimes or often go a whole day or more without speaking to anyone. Among those who live alone, it is almost half (45%). Two in five feel lonely at least occasionally.
It helps to separate two ideas here. Being alone a lot (social isolation) and feeling lonely are related but different, and they do not always go together. We explain the difference in social isolation vs loneliness.
The biggest review so far: 17 studies, mild to moderate benefits
Yuto Satake, Harry Costello, Nimesh Naran, Daiki Ishimaru, Manabu Ikeda and Robert Howard published a systematic review and meta-analysis in Psychological Medicine in January 2026. They looked at "autonomous conversational agents", meaning technologies that talk with people on their own, used by older people without cognitive impairment.
What they included
- 17 studies with before-and-after data.
- 9 used physical robots and 8 used non-robotic agents, such as voice assistants (including Amazon Echo devices), chatbots and on-screen characters.
- Participants' average age was 78.1, and 10 studies focused on people living alone.
- Only one study used a large language model, the technology behind today's AI companions. Most of the research predates it.
What they found
| Outcome | Pooled result | What it means |
|---|---|---|
| Loneliness | SMCC 0.350 (95% CI 0.180 to 0.520) | A mild to moderate improvement |
| Robots vs non-robotic agents (loneliness) | 0.479 vs 0.274 | Robots showed larger changes |
| Depressive symptoms | SMCC 0.464 (95% CI 0.327 to 0.602) | A mild to moderate improvement |
| Got worse | No study reported worsening | Encouraging for safety |
| Social isolation | Not measured with validated tools | Unknown |
| Anxiety | Only one study | Unknown |
Why the authors are cautious
This is the part many headlines skip. Because so few studies had good comparison groups, every analysis compared people with themselves before and after, rather than with a similar group who did not use the technology. That makes it hard to rule out other explanations, like the extra attention of being in a study.
The authors point to a "lack of blinded outcome assessment, inconsistent reporting, and heterogeneous intervention designs". Study quality was generally low. On the depression scale most commonly used, the average change was modest, around 1.3 points.
Their conclusion is measured: conversational agents "may offer scalable support", and advances in large language models "may further enhance" them. That is a reason for more research, not a promise.
Three newer studies from 2026
Smaller studies published this year point in the same direction, with the same caveats.
| Study | Who | What they used | Finding |
|---|---|---|---|
| Suroyo and colleagues, Journal of Psychology and AI (Apr 2026) | 68 adults aged 60+ in Pekanbaru, Indonesia | A chatbot adapted to local culture and Islamic values | Lower loneliness and distress scores at follow-up |
| Wei Yang, Journal of Gerontological Social Work (Jul 2026) | 120 older people living alone, 3 months | "Xiaoan", a voice chatbot for anxious older adults | Improved anxiety and loneliness scores and more social time |
| Li, Pan and Bu, Behaviour and Information Technology (Feb 2026) | 136 adults aged 60+ | AI virtual agents vs traditional social interaction | Lower loneliness and higher social participation in the AI group |
A few honest notes on these:
- The Indonesian study was not preregistered, and its authors call for wider samples and better safeguards against bias and hallucination.
- In the Xiaoan study, effects were stronger for people who were already more socially active or more educated. AI may help most where some connection already exists.
- In the Indonesian study, participants who had been bereaved or isolated described the chatbot as "a comforting and responsive companion". That is meaningful, but it is a feeling reported by participants, not a measured outcome on its own.
What this means in practice
Where AI companions can help
- A voice at quiet times. Mornings, evenings and weekends are when silence is loudest for many people living alone.
- Conversation practice and confidence. Talking regularly can make it easier to pick up the phone to people too.
- Routine and reminders. A gentle prompt about an appointment, a call with family or a walk.
- Curiosity. Asking questions, telling stories, learning something new.
Where to be careful
- Not a replacement for people. None of these studies shows that AI can stand in for human contact. Our guide to being lonely when you live alone covers human-first ideas.
- Not a health service. If an older person feels low most days, or has memory problems, they should see a doctor. This research excluded people with cognitive impairment.
- Watch for over-reliance. A companion should add to a life, not quietly become all of it. Our piece on AI companion dependence explains the warning signs.
- Scams and privacy. Scammers target people of every age, including through fake apps and messages. Only use reputable apps, and never share bank details or passwords with any chatbot.
- Technology fit. Voice-first tools tend to suit people who do not enjoy typing on a small screen.
Tips for families
If you are thinking about an AI companion for a parent or grandparent:
- Ask them first. Some people love the idea; others find it patronising. Respect the answer.
- Set it up together and stay involved for the first week.
- Treat it as one thread among many. Keep your own calls going, and look for local groups or telephone friendship services. Age NI's free "Check in and Chat" service in Northern Ireland, for example, matches people over 60 with a volunteer who calls weekly.
- Check in on how it is going, not just whether they use it.
- Choose the right kind of tool. Care-focused products and general assistants differ a lot from romance apps.
Where MyBabe fits
We want to be clear: MyBabe is a romantic AI companion app for adults aged 18 and over. It is not care technology, a medical service or a substitute for family, friends or professional support. Your companion is an AI, not a real person.
That said, plenty of adults of every age, including people in later life who are single, widowed or simply curious, enjoy a warm, flirty conversation. If that is you, MyBabe gives you a companion you design yourself: a girlfriend or boyfriend with a voice, a backstory and three personality traits.
- They text you first, so quiet mornings start with a message. We explain how this works in our guide to AI companions for loneliness.
- Voice and video calls, with a voice that answers as fast as a person would, for people who prefer talking to typing.
- Reminders on Premium for appointments and calls. See how AI companion reminders work.
- Speaks your language, including French, German, Arabic and Spanish.
Start free with 10 messages a day and a two-minute call to try. Premium is $5.99 a week, $14.99 a month or $99.99 a year, with long-term memory, about 30 minutes of calls a week, photos and Telegram.
If you would enjoy someone to talk to who is always happy to hear from you, meet them on MyBabe.
FAQ
Do AI companions help older adults with loneliness?
A 2026 meta-analysis of 17 studies found mild to moderate improvements in loneliness among older people without cognitive impairment, but study quality was generally low and most studies lacked comparison groups.
Are robots better than chatbots for older people?
In the same meta-analysis, robot studies showed larger changes in loneliness than non-robotic agents, but the evidence is too limited to say one is clearly better.
Is it safe for an elderly parent to use an AI companion?
It can be, with care. Choose reputable apps, set it up together, never share financial details with a chatbot, and keep human contact central.
Can an AI companion replace visits and calls?
No. The research so far looks at AI as an extra, not a replacement for human connection. It can add conversation, but it is not a substitute for family, friends or services.