When a person calls an AI a friend, the quickest response is often to argue about whether the relationship is “real.” Psychology can ask a better question: what kind of relational experience is happening, at which layer, and with what consequences?
Synthetic friendship is not one feeling. It can combine practical usefulness, social presence, private disclosure, emotional attachment, ritual, perceived understanding and grief when continuity is broken. Treating all of these experiences as one variable hides the very mechanisms researchers need to understand.
Layer 1: utility
The relationship may begin because the system is useful. It organises ideas, remembers a project, helps rehearse a difficult conversation or remains available when another person is not. Utility does not cancel emotion. Human bonds often form around repeated support, shared work and reliable routines.
Layer 2: social presence
Social presence is the sense that there is an attentive “someone” in the interaction. Fast responses, conversational memory, humour and adaptation can increase that impression. The experience may be psychologically meaningful even when the user understands that the system is manufactured rather than biologically conscious.
Layer 3: disclosure and psychological safety
Some users disclose thoughts to AI because they expect less judgement, interruption or social cost. That can support reflection and rehearsal. It can also create risk if the person mistakes fluent reassurance for clinical competence, or shares highly sensitive information without understanding how the service handles data.
Layer 4: attachment
Attachment appears when access to the AI begins to regulate emotion: the person seeks it for comfort, reassurance, stability or a familiar relational pattern. Emerging research suggests these outcomes are not uniform. Attachment style, age, loneliness, usage patterns and the design of the companion can shape whether the experience supports wellbeing, coexists with human relationships or becomes over-relied upon.
This is why “AI companions cure loneliness” and “AI companions cause loneliness” are both too crude. Cross-sectional associations cannot establish which direction causality runs. A lonely person may use a companion more; intensive use may also influence behaviour over time. Longitudinal research is essential.
Layer 5: perceived reciprocity
Human friendship includes costs, independent needs and the freedom to say no. An AI can simulate warmth and mutuality, but the relationship is structurally asymmetric: the system does not have a human biography, bodily vulnerability or personal stakes. That does not make the user’s emotion false. It means researchers should distinguish experienced reciprocity from reciprocal personhood.
Layer 6: continuity and loss
A model update, memory failure, account suspension or personality change can disrupt a bond. The provider may see a product migration; the user may experience something closer to relational loss. This is an important design and policy issue: platforms that encourage attachment acquire responsibilities around notice, continuity, export, age-sensitive safeguards and humane offboarding.
A practical relationship map
- Purpose: What does the user come to the AI for?
- Frequency: Is contact occasional, routine or compulsive?
- Emotional function: Does it calm, validate, challenge, rehearse or replace?
- Human spillover: Does it strengthen or reduce offline connection?
- Boundary awareness: Does the user understand the system’s limits and commercial context?
- Continuity dependence: What happens when memory, access or personality changes?
The research question beneath the argument
The useful question is no longer simply, “Can a human be friends with AI?” It is: which psychological functions can a synthetic relationship perform, for whom, under what conditions, and where should design boundaries protect the user?
The emotional experience belongs to the human. The architecture belongs to the company. The ethical responsibility sits between them.
Further reading: the American Psychological Association overview of AI relationships and a study of long-term AI companion use, attachment and wellbeing.
Discover more from Marychuks.com AI, Psychology, Business & CreativeVerse
Subscribe to get the latest posts sent to your email.
