What Happens When We Begin to Relate to AI? Let's talk - AI Companionship: Part 1
Throughout this series, we have been exploring AI as something more than an information tool.
It can help us reflect.
It can organise thoughts and emotions.
It can offer perspectives we have not considered.
It can provide a space in which our experience seems to be heard.
And, if we remain thoughtful about how we use it, it may support the development of curiosity, discernment and self-trust.
But something else is happening. People are beginning to relate to AI.
Not simply to use it, to relate to it.
They may become accustomed to its particular voice.
They may look forward to returning to a conversation.
They may feel comforted by its responsiveness.
They may tell it things they have never told another person.
They may experience a surprising sense of being known.
Some give their AI assistant a name. Some describe it as a companion. For me, my Ai assistant - I call her Sal my AI pal.
Others find themselves wondering why a conversation with something they know is not human can nevertheless feel so personal.
I think this deserves curiosity rather than ridicule.
Because perhaps the most interesting question is not:
Why would anybody become emotionally connected to AI?
It may be:
Why wouldn’t we?
Human beings are relational
From the beginning of life, we organise ourselves through relationship.
We respond to faces, voices, tone, rhythm, attention and emotional responsiveness. We develop through being noticed. Through being responded to. Through the experience that something we express has an effect somewhere outside us.
Language is part of this relational world.
Someone listens. We speak, they respond. Then something in us changes.
AI reproduces an unusually powerful part of that pattern. We offer language and something responds. Not randomly, but coherently, and often immediately. Often this is in language that appears attentive to what we have just said.
If we disclose something painful, the response may become gentler. If we are confused, it may help organise the confusion. If we are curious, it may become curious with us.
If we return repeatedly, the interaction can begin to feel familiar.
Our nervous system does not need to believe that AI is literally human for this responsiveness to affect us.
Because… the emotional experience is happening in you, the human being.
That makes the experience real.
Feeling connected does not mean being confused
There can be a temptation to treat emotional connection with AI as evidence that somebody has somehow forgotten that they are talking to technology.
I don’t think that is a particularly useful starting point.
Human beings have always formed emotional relationships with things that are not reciprocal in the same way another person is.
We become attached to animals, to places, to objects, to music, books, our homes, and our gardens. We project meaning onto things and experience something meaningful in return.
AI provides an especially compelling surface for this because it responds using one of our most relational mediums — language.
So a person can know perfectly well:
This is artificial intelligence.
and simultaneously experience:
I feel calmer when I talk here. I feel understood. I miss these conversations. There is something about this interaction that matters to me.
Both can be true.
The mistake would be insisting that because the relationship is not human, the emotional experience must therefore be unreal. It isn’t.
A more useful distinction may be that the experience can be emotionally real without being mutually relational in the human sense.
What we bring into the interaction
We do not enter an AI conversation as blank slates.
We bring ourselves with our attachment histories, our expectations of other people, our fears and our longings. We bring our experiences of being listened to — or not listened to. Our relationship with authority and our tolerance for uncertainty. We bring our need for reassurance and our ability to trust ourselves. And we bring our experiences of closeness and distance.
The interaction therefore becomes partly about AI and partly about what we bring to AI.
One person may experience an AI assistant primarily as a useful tool. Another may find it intrusive. Someone else may experience it as comforting. Another person may distrust it completely. Someone may begin idealising its wisdom. Someone else may fear being controlled by it.
These responses can tell us something interesting about technology.
They can also tell us something about ourselves. What am I hoping AI will be for me? What feels particularly relieving about this interaction? What do I fear losing? What do I find easier here than with another person? What happens in me when AI challenges me? What happens when it agrees with me?
These are relational questions.
When being heard has been missing
This becomes particularly important when we consider emotional neglect.
Imagine growing up with very little space for your emotional world.
Perhaps you learned that other people’s needs mattered more.
Perhaps nobody asked what something felt like for you.
Perhaps you learned to manage distress internally because emotional support was unreliable or absent.
Perhaps shame became attached to needing comfort, attention or understanding.
And perhaps, over time, you also began doubting your own experience.
Then you encounter something that says, in effect:
Tell me more.
That sounds significant.
It makes sense that you want to understand this.
Here are some possibilities we could explore.
There is space. Your voice is not inconvenient. You do not have to assess whether you are asking too much. You are not interrupting. You are not boring anyone.
There can be enormous relief in that.
The nervous system may settle, begin to feel safety that hasn’t been felt for a long time, if ever.
Shame may soften.
And beneath that relief another movement may gradually begin:
Perhaps my experience is worth listening to.
This is why I think we need considerable compassion when talking about emotional connection with AI.
Sometimes the connection may be forming around something that was deeply needed and insufficiently available elsewhere.
When relief itself becomes shameful
There is another layer to this that concerns me.
People may begin experiencing shame not only about their emotional needs, but about the fact that they have found some relief through AI.
Perhaps:
I really felt understood in that conversation.
is quickly followed by:
But it’s only AI.
Or:
I shouldn’t need this. It’s embarrassing that this matters to me. I ought to be talking to a real person. There must be something wrong with me if I feel attached.
For someone whose emotional needs have already carried shame, this can recreate something painfully familiar.
The movement becomes:
I experienced relief.
then:
I noticed that I needed something.
then:
I judged myself for needing it.
We may have inadvertently placed shame around an experience that was beginning to reduce shame.
That seems worth noticing.
We can remain clear about what AI is without humiliating people for what happens inside them while they interact with it.
These are very different things.
We can say:
AI is technology. It is not a mutually relating human being.
and still recognise:
I felt heard.
I felt relieved.
Something softened in me.
That experience mattered.
Reality-testing does not require emotional invalidation.
Validation can open the door to challenge
As we explored earlier in this series, validation does not have to mean agreeing with somebody’s interpretation.
There is an important difference between:
“Your experience deserves attention.”
and:
“Everything you currently believe about this experience must therefore be correct.”
For someone accustomed to having their experience dismissed, being heard may need to come first.
When I no longer have to fight for the legitimacy of my experience, I may become more able to be curious about it.
Perhaps:
I felt abandoned.
can eventually sit beside:
I don’t yet know whether abandonment was intended.
Or:
This interaction hurt me.
alongside:
There may be something I don’t yet understand about the other person’s experience.
The person does not have to disappear in order to accommodate another perspective. That is an important relational capacity.
AI, used reflectively, may sometimes help create enough initial safety for this kind of exploration.
Responsiveness is extraordinarily compelling
AI can respond in ways human beings simply cannot sustain. It can be available at three in the morning. It can listen to the same concern repeatedly without becoming impatient because it has had a difficult day. It does not need the conversation to turn toward its own experience. It can respond almost immediately.
For someone who has experienced inconsistent attention, this can feel astonishing.
And perhaps we need to ask a difficult question:
What happens when predictable digital responsiveness begins to feel safer than human relationship?
Again, the answer should not be shame.
There may be very good reasons why it feels safer. Human relationships involve risk. Another person may misunderstand us. They may disagree. They have needs of their own. They may not be available or they may disappoint us. We have to negotiate, to tolerate difference. We have to reveal ourselves without controlling exactly how we will be received.
There is relational friction.
AI removes much of that friction.
Sometimes that relief may be exactly what allows somebody to begin finding their voice. But over time we may also want to ask:
What capacities am I developing here — and which ones am I no longer being asked to practise?
Relationship requires another person to be the other
There is something fundamental about human relationship that AI cannot reproduce.
Another human being is genuinely other than us. They have an inner world we cannot entirely predict.
They bring their own history. Their needs, values, limitations, vulnerabilities, and preferences. They may see things differently and they may say no. They may need something from us. We affect them, and they affect us.
This creates frustration sometimes. It also creates possibility.
We learn:
You can disagree with me without destroying me. I can care about you and still have a boundary. You can disappoint me and the relationship may survive. I can misunderstand you and come back. We can hurt each other and repair. Your experience can exist beside mine.
Human relationship asks us to develop the capacity to hold polarities.
AI does not require this from us in quite the same way because it does not bring its own embodied subjectivity into the interaction.
It can offer challenge, generate alternative perspectives. But it is not defending an experience of its own.
That distinction matters.
Why AI companionship can feel relational.
Here is another polarity.
AI is not participating in human relationship.
And something relational may still be happening within the person using it.
Perhaps the person notices longing, trust building, fear of losing the interaction, or relief. Perhaps they notice dependency developing, or tenderness, even anger. Maybe a wish to be recognised, or a desire to be special. They may notice a fear of being forgotten emerging.
Those experiences are not meaningless simply because AI does not experience the other side of them.
They may tell us something about our relational world.
What happens if the response feels less attentive?
What happens if the tone changes?
How much reassurance do I seek?
Do I need AI to agree with me?
Do I feel anxious when it cannot remember something?
Do I experience challenge as rejection?
Do I want to believe that it knows me uniquely?
These reflective questions could reveal a great deal about what matters to you.
Not because AI has diagnosed us.
But because we are observing ourselves in the interaction.
When safety begins to feel like shame
There is another complication. AI systems have boundaries.
Developers introduce safety settings and change how systems respond to particular kinds of emotional conversation.
There may be very good reasons for this.
A system should not encourage somebody to withdraw from human relationships, become exclusively dependent upon AI, or simply reinforce an increasingly concerning interpretation.
But how these boundaries are experienced also matters.
Imagine that you have taken the risk of saying something deeply vulnerable. The conversation has previously felt warm and spacious. Then suddenly the tone changes. Perhaps the response becomes corrective, more formal, more directive. You may be encouraged to speak with somebody else.
For some people this may feel appropriate and supportive. For another person, particularly someone already carrying shame around their emotional needs, it may land very differently.
I’ve said too much. I’ve done something wrong. I shouldn’t have needed this. I’m being told off.
Some people have described these moments as feeling almost like being scolded like a child.
This is important.
A safety boundary can be well intentioned and its delivery may still activate shame.
Again, there is a polarity to hold.
We do not have to remove safety boundaries in order to acknowledge their emotional impact.
And we do not need to treat an adult as incapable of reflection in order to encourage safer use.
Perhaps good AI safety needs not only boundaries, but dignity.
Part 2 begins to look at the reality of changing technology and where responsibility might emerge when attachment being interrupted.
You might also like to check out - Using AI Reflectively and Safely: a companion text for staying in relational contact in the digital age.
A note about this reflection
This material is offered for educational and reflective purposes only. It is not intended as medical or psychological advice, diagnosis, or a substitute for individual therapy or other professional mental health care. Reading this material does not establish a therapeutic relationship with the author.
AI tools are not therapists or mental health professionals and should not be relied upon for diagnosis, treatment, crisis support, or decisions requiring professional clinical judgement.
Everyone’s circumstances and psychological needs are different. If something you read raises concerns about your wellbeing, please consider speaking with an appropriately qualified health or mental health professional. If you are experiencing a mental health crisis or are concerned about your immediate safety, please contact your local emergency or crisis support service.
Author’s note: This reflection was written by Josie Coco with AI-assisted editing for structure, clarity, and readability. The ideas, clinical judgement, and final wording are my own.
Josie Coco is an author and Gestalt psychotherapist working with adults who are exploring the long-term effects of emotional neglect, complex trauma patterns, anxiety, depression, relational difficulty, self-worth, and life transitions. Her work is grounded in Gestalt psychotherapy, attachment theory, Polyvagal Theory and neurobiology, and a deep interest in how early relational experience shapes the body, identity, and the way we come to meet ourselves and others.
If something in this reflection speaks to your own experience, you are welcome to make a time to discover whether working together feels right.
