People can already form emotional attachments to AI companions that exist only as apps or chatbots. Pairing those systems with a capable humanoid robot is technically plausible over the longer term, but current evidence gives no basis for predicting that machines will replace human romantic partners.
The more immediate question is narrower: what happens when conversational systems that remember preferences and respond at any hour become more persuasive, more personal and eventually more physical?
That matters because the market is already forming around software companions. A humanoid body could add mobility and touch, but it would also turn the product into a far more expensive combination of hardware, cloud software, maintenance and personal data.
AI companionship already exists without a robot body
A 2026 review in Computers in Human Behavior Reports describes “synthetic relationships” as ongoing bonds with AI companions designed to simulate human social connection. The review identifies constant availability, adaptability and emotional responsiveness as possible attractions, while also raising risks of emotional over-reliance, distorted expectations and privacy loss.
The authors argue that these systems may have a place when they complement human relationships. That is especially relevant in a world where, according to the World Health Organization, about one in six people experiences loneliness.
For a person who has recently lost a partner, moved away from friends or spends long periods alone, an AI companion may not be competing with an ideal human relationship. It may be competing with silence. That does not settle whether it improves a person’s long-term social life, but it explains why the products appeal to some users.
What the research says about intimacy with AI
Research is beginning to separate different types of companionship instead of treating all human-AI relationships as the same. In a 2026 survey of 447 people, researchers found stronger willingness to consider AI in cognitive or conversational roles, such as a friend, mentor or therapist, than in romantic, sexual or caregiving roles.
The survey also found that embodiment affected preferences. Respondents tended to favor virtual agents for cognitive roles, while physical robots were more appealing where touch or practical activity was part of the role. It measured stated attitudes, not what people would do after living with a robot, so it cannot forecast consumer demand for an artificial partner.
Evidence on loneliness is also mixed. A 12-month study of more than 2,000 adults in four Western countries found that increased social-chatbot use predicted higher emotional isolation on one measure. The researchers urged caution because parts of the analysis were exploratory, and a broader measure showed that lower social connection predicted more later chatbot use, without showing that chatbot use reduced connection.
In a separate preregistered experiment, 296 first-year university students spent two weeks either messaging a supportive chatbot, texting a randomly assigned peer or writing a brief daily journal entry. The human-peer group showed a greater reduction in loneliness. The chatbot did not outperform journaling on that outcome.
Humanoid robots are still built for work
The robotic half of the idea remains a major limitation. Boston Dynamics’ production Atlas program, for example, is aimed at industrial tasks. It has to move safely, perceive its surroundings, pick up objects and operate reliably in a workplace before it can earn its place in a factory.
A household companion would need those physical abilities plus safe interaction around people, conversational reliability, long-term memory and a way to function in homes that have not been designed for robots. It would also have to be affordable enough to move beyond a niche product. No company has demonstrated that full package in a consumer relationship robot.
Our earlier coverage of why hands-on jobs remain less exposed to AI examined the gap between a software system that can advise on a task and a machine that can complete it safely in the physical world. Human companionship would add a more demanding layer: trust, emotional dependence and consent.
The business model may matter as much as the hardware
A conversational companion with a body would not be a conventional appliance. Its personality, memory and behavior would likely depend on a service run by a company. A user might own the robot while relying on a provider for model updates, cloud memory, voices, safety settings and repairs.
That creates commercial questions with no equivalent in a human relationship. Can a provider alter a companion’s behavior in a software update? Would its most intimate functions sit behind a subscription? Can a user export years of conversations if the provider closes or changes its terms?
Those questions are already relevant for app-based companions. A humanoid form would make them harder because the product could become part device, part service and part repository of a person’s private history.
Human partners are independent people with their own interests, limits and ability to disagree. A commercial AI companion can be designed around one user’s preferences. That may make it comforting or convenient, but it cannot establish whether a relationship between two independent people has been replaced. The technology may find its first durable market in companionship and assistance for people who want it, while human relationships remain a different category altogether.
