Companion AI that listens, remembers, and responds with genuine social presence — built on what the science of conversation actually tells us.
Purpose
People are increasingly turning to conversational AI not just for answers, but for company. Done carelessly, that is a risk. Done well — grounded in emotion research and honest about its limits — it can genuinely support people who are isolated, grieving, or simply want to be heard.
AI Companion is 42.AI's research into that possibility: agents that hold context over time and respond with social presence, designed from the start around consent, transparency, and psychological safety.
Try it: a companion conversation
A short, scripted demo — nothing you tap is sent or stored. It only hints at what a warm, remembering companion might feel like.
Companion demo · no data collected
Hi — it's good to see you again. How are you feeling today?
Illustrative only. A real companion would be built with consent, privacy, and safeguards at its core.
Research & writing
Foundations
The first chatbot revealed how easily we bond with machines
In 1966, Joseph Weizenbaum built ELIZA, a program that imitated a therapist by reflecting users' statements back as questions. It understood nothing — yet people confided in it, and some became convinced it truly cared. Weizenbaum was so unsettled that he spent much of his later career warning about the effect.
That phenomenon — now called the "ELIZA effect" — is the central design challenge for any companion AI. The tendency to attribute understanding and feeling to a responsive system is powerful and immediate. AI Companion treats it as a responsibility: presence must be matched with transparency about what the system is and is not.
Reference
Weizenbaum, J. (1966). ELIZA—a computer program for the study of natural language communication between man and machine. Communications of the ACM, 9(1), 36–45. https://doi.org/10.1145/365153.365168
Evidence today
Can talking to an AI actually ease loneliness?
A 2024 study surveyed more than a thousand student users of the companion app Replika. The participants were markedly lonely — and many described the AI as a friend, a sounding board, even a kind of therapist. Strikingly, a small share reported that the companion had interrupted their suicidal thinking.
The authors are careful not to overclaim, and neither are we: the findings are correlational and the population is specific. But they establish something important — companion AI is already meeting real emotional needs at scale. That makes getting the design, the safeguards, and the ethics right an urgent problem, not a hypothetical one. It is exactly the problem AI Companion exists to work on.
Reference
Maples, B., Cerit, M., Vishwanath, A., & Pea, R. (2024). Loneliness and suicide mitigation for students using GPT3-enabled chatbots. npj Mental Health Research, 3(1), 4. https://doi.org/10.1038/s44184-023-00047-6