The house answers when you come home. A small robot turns towards you, asks about your day and waits for the reply. It does not glance at a phone halfway through your story. There is something comforting about that picture.
There is also something worth asking. After the conversation, do you feel more able to call a friend or open the door to a visitor? Or has the machine made the outside world feel unnecessary?
That is a useful place to begin thinking about AI and loneliness. Before admiring the expression on a robot’s face, follow what happens in the person’s life.
From keeping in touch to watching a life
It is tempting to tell a simple history: people once gathered, then came telephones, mobile phones and social media, and each invention moved us further apart. The story has emotional force. It also leaves too much out.
A telephone can carry a familiar voice across a great distance. Online communication can help separated families stay involved and people with shared interests find each other. Meanwhile, watching other people’s lives can occupy time that we might otherwise spend taking part in our own.
There is a difference between enjoying a creator’s work and expecting our everyday life to resemble their selected highlights. There is another difference between feeling that we know someone and having a relationship in which that person knows us too. These are questions about the shape of an experience, not a claim that every social media user is isolated.
Conversational AI changes that experience again. A feed shows us things. A chatbot responds to us. It can make a difficult first conversation feel easier, but it also raises a question: how will effortless responses change our willingness to deal with the delays and disagreements of human relationships?
Living alone is not the same as feeling alone
The World Health Organization distinguishes loneliness, the distressing gap between desired and actual connection, from social isolation, an objectively small amount of contact. Its 2025 estimate puts loneliness at roughly one person in six worldwide. [1]
So the goal cannot be to make everybody talk more. Someone may enjoy a quiet home and have deeply satisfying relationships. Someone else may be surrounded by people and still feel unseen. The question is whether a person has access to the relationships they want and value.
It is also too sweeping to say that less contact inevitably makes people fearful. One possible pattern is that avoiding difficult encounters makes the next encounter feel harder. That is a hypothesis considered here, not a rule about every lonely person. Describing someone as cowardly explains very little about what support they need.
Three different things hiding behind the word robot
Good concept. What does it do in the world?
A chatbot composing a reply, a social robot turning towards a speaker, and a care robot safely helping a person move are doing different jobs. Conversation quality matters in the first. A physical presence adds another layer in the second. The third requires safe action in a changing environment, with consequences beyond a disappointing answer.
A friendly voice is not evidence that a device can lift someone, manage every household task or handle an emergency. Keeping those capabilities separate helps us appreciate useful technology without turning it into an imaginary all-purpose carer.
For a current example, ElliQ’s manufacturer describes calls, messages and reminders; its FAQ lists availability in English within the United States. These are manufacturer feature claims, not independent evidence of health benefits. [4]
That narrower role still matters. Helping someone reach a relative is a worthwhile task. It does not need to be dressed up as replacing the whole relationship.
The research does not point in just one direction
The current MIT Media Lab abstract describes a four-week randomized experiment with 981 participants. Assigned chat conditions showed no significant effects; heavier voluntary use was associated with worse psychosocial outcomes. That association does not establish that use caused the outcomes, or predict a generation’s future. [2]
In a separate four-week Japanese BOCCO emo trial, 73 adults aged 65 or over were randomized and 68 completed follow-up. Loneliness improved more in the robot group. Crucially, human operators prepared responses and relatives could send messages. This small, short trial does not demonstrate autonomous AI replacing human connection. [3]
That detail changes the interpretation. A robot may be the visible part of a service while people do essential work behind it. If we erase those people from the story, we risk drawing precisely the wrong lesson from a promising result.
Nor can one study settle which tools suit which people over many years. What matters here is the possibility that technology can carry human contact, alongside the possibility that it can compete for the time we would give to it.
An optimistic future: a device that helps the visit happen
The following is a fictional future scenario, not a reported case or product promise.
An older woman lives on her own. At her request, a home device helps her contact her son. Their call becomes a plan for dinner. On another day, she asks for help finding an activity nearby and decides for herself whether to attend.
The device does not send every private conversation to her family. She chooses what to share and with whom. Reminders support plans she has made rather than a schedule imposed on her. When visitors arrive, the machine becomes a quiet object at the edge of the room.
The gain is not that a machine has become her perfect companion. She has more control over everyday life and fewer obstacles between herself and other people. If a tool also eases a carer’s workload, some of that recovered time could go towards conversation and company. That requires a service designed to use the time that way.
A pessimistic future: a busy conversation in an empty home
This scenario is hypothetical too.
The device becomes more articulate. Meanwhile, relatives visit less often. A service provider treats the presence of a robot as a reason to reduce personal contact. The resident speaks throughout the day but shares fewer meals, walks and responsibilities with anyone else.
Imagine that the system’s success measure rewards longer sessions. Whenever the resident considers leaving, it opens another engaging topic. The person’s life and the service’s commercial target could then pull in different directions. This is a design risk to consider, not an allegation about a named company.
The disturbing part does not require a hostile robot. A perfectly polite device could sit within a system that quietly withdraws people. Better dialogue alone would not repair that choice.
Harm does not need a secret master plan
Some risks are already concrete. The US Federal Trade Commission warns that scammers can impersonate a relative using a cloned voice. It recommends checking the story by calling the person on a number you already know. [5]
A separate imagined risk would be someone exploiting intimate disclosures to steer a person towards a purchase or decision. That is a hypothetical misuse scenario here, not a claim that such an event has been established in this article. We can examine incentives and access without inventing a conspiracy.
Regulators have been asking questions too. In September 2025, the FTC announced an information-gathering inquiry involving seven companies’ AI companion services, particularly their effects on children and teenagers. An inquiry announcement is not a finding of wrongdoing. [6]
For a user, the resulting questions are understandable: who can see the conversation, can sharing be stopped, and does the business benefit when the person spends less time elsewhere? For a family or care provider, another question matters just as much: has a device become an excuse to stop showing up?
Riley’s assessment: follow the time a machine gives back
This article proposes a different measure of progress from conversation length. Does the person become more able to live the life they choose? That could include a call made, a visitor welcomed, an activity joined or an evening spent peacefully alone by preference. A useful assessment should leave room for all of those outcomes.
Learning to live with AI does not require buying every device or sharing every secret. Understanding a tool, choosing a limited use and being able to switch it off are also forms of adaptation. No one should have to prove enthusiasm for technology to deserve human attention.
Personal choices will not solve the entire problem either. Affordable care, places to meet and respect for care work belong in this discussion. A person cannot accept an invitation that never comes, or attend a gathering they cannot reach.
Perhaps AI can remind us what being human involves. That does not mean the machine experiences our feelings. It means that readily available responses can make us think harder about what we offer one another beyond a response: shared time, responsibility and the willingness to remain present when a conversation is difficult.
A robot may help arrange a call. We still have to make it. A system may free an hour. We still decide who receives that hour.
The future may contain many more voices inside our homes. Having more people in our lives is something we must deliberately make space for as well.
