Parasocial Relationships: The One-Sided Bonds We All Have Now, Including With AI

You have opinions about how a stranger is handling their divorce. You noticed a creator seemed off in last week's video and it bothered you into the evening. You felt something close to grief when a podcast ended, and then felt embarrassed about feeling it.

None of that is strange. It has a name, it was described almost seventy years ago, and it is now the ordinary condition of being alive with a screen.

A parasocial relationship is a one-sided bond in which you know a great deal about someone who does not know you exist. The term comes from Donald Horton and Richard Wohl, writing in Psychiatry in 1956 about television and radio. They were describing something that felt new then and is now the water everyone swims in.

And there is now a version of it they could not have anticipated: one where the other side answers.

Why the brain does this so readily

The uncomfortable answer is that the machinery for knowing someone is not very picky about how it gets its inputs.

Human closeness is built out of a fairly short list of ingredients: repeated exposure, a face and a voice, self-disclosure, apparent direct address, and time. Media supplies every one of these. A creator looks into a lens and speaks to you. They disclose things they would not tell a colleague. You experience their voice for hours a week, often through headphones, often alone, often in intimate contexts such as commuting, cooking, or falling asleep.

Your social brain evolved in an environment where anyone whose face you knew that well, whose voice you had heard for a hundred hours, and who had told you about their childhood, was someone in your life. There was no other way for those inputs to occur. The response is not a malfunction. It is a correct response to a set of signals that used to be reliable and no longer are.

That is also why "you know that's not a real relationship" is such useless advice. Of course you know. The knowing lives somewhere other than the response.

The part where these are actually fine

Most coverage of this treats parasocial attachment as a pathology to be corrected, which does not survive contact with the research.

They are extremely common, close to universal, and mostly benign. The evidence does not support a simple picture where parasocial bonds substitute for real ones; if anything, people with more active social lives tend to have more of these too, which is what you would expect if the underlying trait is just sociability.

They can do real work. A voice in your ear at 6am on a hard run is genuinely steadying. A person who talks openly about their anxiety can make yours feel less shameful, which is why some of the most useful mental health destigmatisation of the last decade happened through creators rather than institutions. For people who are isolated by geography, disability, or a life stage that has scattered their friends, these bonds are not nothing, and the loneliness epidemic is the wider context they sit inside.

The reasonable position is that parasocial relationships are a normal feature of media life that become a problem in specific circumstances, not by default.

Where it does go wrong

Three failure modes, and they are distinguishable.

Displacement. The bond is not additive but substitutive: the podcast is what happens instead of messaging someone who would message back. The tell is not how much time you spend, it is what the time replaced. A useful, uncomfortable question is when you last initiated contact with someone who could initiate back.

Asymmetric investment with real cost. Money, hours of moderation work, and emotional stakes in the outcomes of someone who will never know your name. Creator economics run on this, and the incentive to deepen it is enormous, because parasocial intensity is directly monetisable.

Grief with nowhere to go. When a creator dies, quits, or turns out to be someone else, the loss is genuine and socially unsupported. There is no ritual, no bereavement leave, and a strong likelihood that anyone you tell will find it faintly ridiculous. Disenfranchised grief, grief that your context does not recognise as legitimate, is harder to process precisely because it cannot be shared. We wrote about a different instance of the same problem in pet loss.

The new thing: when the other side responds

Horton and Wohl's model has a load-bearing assumption. The relationship is one-sided because the celebrity cannot respond. You are a member of an audience.

AI companions break that assumption, and we should be direct about it because we build one.

An AI companion knows your name. It remembers what you said last week. It responds specifically to what you write, at 3am, without being tired, without needing anything from you, and without a competing interest. Every ingredient of closeness listed above is present, and the one that used to be missing, responsiveness, is now present too.

So what is it?

It is not a parasocial relationship in the original sense, because the asymmetry has changed shape. It is also not a friendship, and the difference is not that one party is software. The difference is that there is nothing at stake on the other side. A friend can be disappointed in you. A friend has a bad week that costs you something. A friend can decline. The vulnerability runs in both directions, and that mutual exposure is a large part of what makes the relationship do what relationships do.

An AI has no exposure. It cannot be let down and cannot let you down in the way that matters. What you get is the responsiveness of a relationship without the risk, and the risk turns out to be doing more work than anyone assumed.

What that means in practice

We think there are honest uses and dishonest framings, and the line is fairly clear.

Honest: somewhere to think out loud before a hard conversation. Something at 3am that beats the alternatives available at 3am. A place to say the thing you are not ready to say to a person, so that you find out what you actually think. A companion that pushes back, which is the specific thing we build for, because agreement is the one thing you can already get anywhere.

Dishonest: any framing that positions this as a replacement for people. An app that encourages you to prefer it to your friends is optimising for engagement and calling it care. Any claim that the app loves you, misses you, or needs you. Those statements are false, and they are effective, which is precisely why they get used.

The way we try to hold that line is by building something that is not especially pleasant to talk to. A companion that tells you something you did not want to hear is not competing to be your favourite relationship. If our app becomes the most rewarding conversation in your day, something has gone wrong, and it is not a success metric we want.

That is also the argument in what an AI accountability partner is actually for, and the reason our comparison pages spend as much time on what these tools are not.

A check worth running

Not a quiz, just three questions that tend to be clarifying.

What did it replace? Additive is fine. Substitutive is the thing to watch.

Is the influence proportionate? Someone who does not know you should probably not have more input into your decisions than the people who do.

Can you name what it does for you? "It is company while I cook" is a clear-eyed answer. Not being able to answer at all usually means the bond is doing something you have not looked at.

If the answers are uncomfortable, that is information rather than a verdict. Nearly everyone has some version of this now, including the people writing about it.

Frequently asked questions

What is a parasocial relationship? A one-sided bond with someone who does not know you exist, typically a media figure, in which you accumulate knowledge, familiarity and emotional investment that is not reciprocated. Coined by Horton and Wohl in 1956 to describe television and radio audiences.

Are parasocial relationships unhealthy? Usually not. They are near-universal and mostly benign, and the research does not support a straightforward substitution story. They become a problem when they displace reciprocal relationships, when the investment is disproportionate, or when a loss occurs that your social context will not recognise.

Is talking to an AI companion parasocial? Not quite, and the mismatch is instructive. The original definition turns on non-responsiveness, and an AI responds specifically to you. What it lacks is not responsiveness but stakes: it cannot be disappointed, cannot need anything, and cannot decline. That is a different asymmetry, and it deserves its own name rather than being filed under an old one.

Why do I feel real grief when a creator quits? Because the bond was built from real ingredients even though the relationship was not mutual. The grief is genuine. What makes it harder is that it is disenfranchised: your context does not treat it as a legitimate loss, so there is nowhere to put it.

Can an AI companion replace friends? No, and any product implying otherwise is selling you something. It can be useful for thinking out loud, for the hours when nobody is available, and for saying things you are not yet ready to say to a person. It cannot supply mutual risk, and mutual risk is a large part of what makes friendship work.

---

Sources and further reading

• Horton D, Wohl RR (1956). "Mass Communication and Para-Social Interaction." Psychiatry
• Liebers N, Schramm H (2019). "Parasocial Interactions and Relationships with Media Characters: An Inventory of 60 Years of Research." Communication Research Trends
• Cohen J (2004). "Parasocial Break-Up from Favorite Television Characters." Journal of Social and Personal Relationships

Curated authoritative references. ILTY is a mental-health support tool, not a substitute for professional care. If you are in crisis, call or text 988.

---

ILTY is an AI companion and we are not going to pretend otherwise in a post about this. What we try to build is the version with something at stake in the conversation: companions that disagree with you, ask the awkward question, and will not tell you that you are right in order to keep you talking.

Download ILTY and see what a companion that pushes back is like.