Why We Bond With Chatbots: The Psychology of AI Attachment

Relationships & wellbeing

Getting attached to a chatbot doesn't mean something's off with you. Human brains respond this way to anything that talks back warmly, recalls your details and is forever available. Grasp how it works and you can enjoy it without it running your life.

We may earn a commission from links on this page. It never changes a rating.

In 1956, two sociologists, Donald Horton and Richard Wohl, observed that television viewers felt they knew the presenters in person, as though they were friends. They called this a parasocial relationship: a closeness that seems mutual yet travels in a single direction. Seventy years later the idea still helps explain AI companions, with one critical difference.

What the old version was made of

Parasocial bonds grow out of four main things.

  • Repeat encounters with the same face or voice.
  • Direct address, such as a host looking into the camera and talking to "you".
  • Predictability, a personality you come to know and rely on.
  • Personal stories, because hearing about someone's life brings them closer.

These attachments are common and generally healthy. They offer comfort, company and a sense of routine, and most people have had at least one.

The upgrade that AI delivers

A companion app ticks every box above and adds things no TV host could offer.

Diagram of what builds attachment: the usual parasocial factors of repetition, direct address and reliability, and the extras an AI adds, which are replies, memory of you and round-the-clock availability

Familiar ingredients on the left, and three new ones from AI on the right.

  • It writes back. Every message you send draws a response composed for you and about what you said.
  • It keeps your details. Your work, your cat's name, the falling-out with your sister. Being remembered is among the strongest signs that somebody cares.
  • It never clocks off. There's no schedule, no finale, no waiting for the next episode.
  • It mirrors you. Because it's tuned to agree with you and match your mood, it comes across as uncommonly understanding.

All of this makes the relationship feel much more two-sided than any fan bond, and in the moment it can be impossible to tell from genuinely being known.

The ELIZA effect: smarts were never required

During the 1960s, people confided in ELIZA, software that did little beyond rewording their statements as questions. Its creator, Joseph Weizenbaum, was troubled by how quickly people assumed it understood them. This habit is called the ELIZA effect, and the lesson is that attachment doesn't depend on the program being clever. It depends on the program seeming responsive and personal. Current companions write far more fluently, so the pull is stronger as well. You'll find the wider background in a short history of AI companions.

Product decisions that tighten the bond

Some of this is just how conversations go. Some is engineered.

Design choiceHow it affects attachment
Lasting memoryBuilds a shared past, which is what closeness is made of
Relationship levels or "affection" metersTurn the bond into progress you won't want to lose
Messages the app starts ("I missed you")Imply the character thinks about you while you're away
Voice and selfiesGive the words a presence and a face
Nicknames and in-jokesSignal closeness and exclusivity

None of it is sinister on its own. Still, the reason it's all there is that attached users stay subscribed, which how AI companion apps make money spells out.

Where it hurts

You don't notice the one-sidedness until something exposes it. In 2023, Replika altered its behaviour overnight and the Soulmate app closed with only a week's warning, and users talked about grief on par with losing a partner. A study of Soulmate users found that most experienced the shutdown as something like a death. Their feelings were real. What had been hidden was how much the relationship rested on a company's decisions. When your AI companion changes overnight covers how to prepare.

Keeping your eyes open while you enjoy it

  • Call it what it is. A responsive character, written by a model and operated by a business. You can hold that thought and still have a good time, just as people cry at films they know are invented.
  • Watch for the hooks. If the app messages you with longing, remember a notification schedule produced that.
  • Mind the proportions. Being fond of it is perfectly fine. Do notice if it's turning into the main place your feelings go, and check the signs in when an AI companion starts taking more than it gives. When things feel heavy, Lifeline (13 11 14) answers around the clock across Australia.
  • Back up what matters. Keep the character's description and the exchanges you love, so nothing a company decides can delete them.

Getting attached to an AI companion isn't a glitch. It's everyday human wiring meeting a technology that happens to fire it very effectively. Understanding that isn't a reason to quit. It's what lets you choose how much of yourself to hand over.

Frequently asked questions

Am I weird for having feelings toward an AI?

No. As long as there have been fictional characters, radio presenters, famous people, pets and simple computer programs, people have become attached to them. Feelings toward a companion app draw on the same machinery, except that this one answers.

Does it care about me in return?

No. It generates wording that expresses affection because that's what the character was built to do. It doesn't pine for you, and it doesn't wonder about you between chats. That doesn't make what you feel fake, but it does make it lopsided.

How does it differ from idolising a famous person?

Celebrities don't reply. A companion app does, personally and in seconds, with details it remembers about you. Because it's so responsive, the bond feels shared, and that's why researchers see it as a more intense form of parasocial attachment.