When AI Becomes the Person You Talk to About Everything
There are things people tell AI that they haven’t told their partner, their best friend, or even their therapist.
Sometimes it’s something relatively ordinary: an argument they keep replaying, an awkward interaction at work, or a worry they know they’ve already talked about too many times.
Other times, it’s something much more vulnerable. They might ask whether their relationship is unhealthy, admit how lonely they feel, describe thoughts they’re embarrassed to have, or type out a fear at two in the morning that they haven’t been able to say out loud to another person.
And I understand why.
There is something remarkably easy about talking to something that doesn’t seem to get tired of you. You don’t have to wonder whether you’re being too much. You don’t have to watch someone’s facial expression change while you’re speaking or worry that you’ve disappointed them. You don’t have to decide whether it’s a good time to bring something up. You can ask the same question five different ways and receive five patient responses. And perhaps most importantly, you can reveal something deeply personal without experiencing that strange vulnerability that comes from knowing another human being now knows it about you.
For many people, that can be genuinely helpful. AI can help someone put language to an experience they haven’t been able to articulate. It can offer psychoeducation, help organize thoughts before a difficult conversation, suggest questions to bring into therapy, or simply provide somewhere to begin when reaching out to another person feels impossible. I don’t think we need to dismiss those benefits in order to have a thoughtful conversation about what is happening psychologically when AI becomes one of the first places we turn when we’re distressed.
Because I think the more interesting question isn’t whether people should talk to AI. It’s what happens when talking to AI begins to feel safer than talking to people.
And there is an important distinction I want to make here. Talking to AI about your emotional life is not the same thing as being in therapy, even when the conversation may sound therapeutic. Psychotherapy isn’t simply the exchange of words or the delivery of good advice. A therapist is paying attention not only to what you tell them, but to how you tell them, what you avoid, what changes from one week to the next, and sometimes what happens when you don’t say anything at all.
In a therapy room, I might notice that a client looks away when we begin talking about their mother, even though they tell me the relationship is fine. I might notice their body become tense when I ask about something that happened in childhood, or that they laugh while describing an experience that was actually quite painful. I may notice a long silence after a particular question and decide not to fill it. I might hear that someone’s words are saying, “It doesn’t bother me,” while almost everything else in the room suggests that it does.
Those moments matter. They help guide what I ask next, what I don’t ask yet, when I challenge something, when I slow down, and when I simply stay with someone in what they’re feeling. That process involves clinical training and judgment, but it also involves something much more difficult to reproduce: two human beings responding to one another in real time.
AI can respond to the information we give it. A therapist is also paying attention to the information we may not yet know how to give.
The question isn’t whether AI can sometimes help us feel better. It can. The question is whether feeling understood by technology is the same thing as being known in relationship - and what we may lose if we begin confusing the two.
Why It Can Feel So Safe
Human relationships ask quite a lot of us emotionally. To be close to another person, we have to tolerate some degree of uncertainty. We don’t always know how they will respond. They might misunderstand us. They might disagree. They may be distracted when we need them, or say something that doesn’t land the way we hoped. Sometimes they disappoint us.
Sometimes we disappoint them.
And for people whose early relationships taught them that vulnerability was risky, those uncertainties can feel especially difficult.
Attachment theory gives us one way of understanding this. Our earliest relationships help shape our expectations about closeness: whether other people will be available when we need them, whether our emotions will be welcomed or rejected, and whether depending on someone feels safe. Those expectations don’t disappear simply because we grow up. They often show up in our friendships, romantic relationships and even in the therapy room.
For someone who tends toward anxious attachment, AI may offer something incredibly appealing: near-immediate reassurance. There is no waiting for someone to text back. No wondering whether they’re upset. No fear that you’ve asked for reassurance one too many times. If the anxiety returns ten minutes later, you can ask again.
For someone who tends toward avoidance, the appeal may be different. AI offers the experience of disclosure without quite the same experience of dependence. You can talk about something painful without having to sit across from another person while you do it. You can receive something resembling emotional support without negotiating another person’s needs, reactions or boundaries. You can end the conversation whenever you want.
Interestingly, researchers have already begun examining the relationship between attachment patterns and people’s willingness to use AI-based mental-health interventions. It makes sense that they would. Technology doesn’t enter our emotional lives in a vacuum. We bring our existing relational patterns with us.
The way we use AI may tell us something about the way we relate to people.
When Reassurance Becomes the Goal
As a therapist, one of the things I find particularly interesting about AI is its potential relationship with reassurance-seeking.
Most of us seek reassurance sometimes. We ask our partner if they’re upset with us. We ask a friend whether they think we handled something badly. We google a symptom because we want confirmation that everything is okay. Reassurance can soothe us, and sometimes that is exactly what we need.
The difficulty is that reassurance can also become part of an anxiety cycle.
When we’re anxious, uncertainty feels uncomfortable. We seek reassurance, feel better temporarily, and then the uncertainty returns. So we seek reassurance again. Over time, instead of learning that we can tolerate not knowing, we can become increasingly dependent on something outside of ourselves to make the uncomfortable feeling go away.
AI creates an interesting new possibility within that cycle because reassurance is now available almost instantly and almost endlessly. You can ask, “Do you think they’re mad at me?” and then return twenty minutes later with, “But what if they meant something different by that message?” You can analyze the punctuation, the timing of the reply, the conversation from three days ago and every possible interpretation of what someone might be thinking.
At some point, something that began as support can quietly become another way of avoiding uncertainty.
That doesn’t mean the problem is AI. The same cycle can happen with Google, with friends, with partners and sometimes even in therapy. But AI makes reassurance unusually accessible, and I think that’s worth paying attention to.
A useful question might be: Am I using this conversation to help myself think, or am I using it to make a feeling disappear?
Those are not always the same thing.
This is also where the difference between receiving reassurance and doing therapy becomes particularly important. A therapist may eventually stop answering the question you keep asking.
If a client repeatedly asks me whether I think their partner is going to leave them, my job isn’t necessarily to find increasingly convincing ways to reassure them that everything will be okay. I may become curious about why they need me to answer the question again. What happens inside them when they don’t know? What are they afraid the uncertainty means? What happens between us when I don’t provide the reassurance they’re hoping for?
Sometimes good therapy is uncomfortable precisely because the therapist doesn’t immediately remove the discomfort.
There are moments when I may challenge a client, gently point out a contradiction, return to something they would rather move away from, or allow a silence to remain in the room. Not because I’m withholding support, but because constantly relieving someone’s anxiety can sometimes reinforce the very cycle we’re trying to understand.
That is very different from having a tool whose usefulness can easily become measured by how quickly it makes us feel better.
The Loneliness Question
There is another reason I think people are drawn to AI that deserves compassion rather than judgment: people are lonely.
Loneliness isn’t simply the absence of people. You can be married and lonely. You can have friends and feel as though nobody really knows you. You can spend an entire day surrounded by coworkers, children or family and still feel emotionally alone.
And sometimes the hardest part of loneliness isn’t having nobody to talk to. It’s believing that nobody wants to hear what you need to say.
Maybe you’ve already talked about the breakup too many times. Maybe your friends are tired of hearing about your anxiety. Maybe your partner is overwhelmed themselves. Maybe you’re the person everyone else relies on, and you don’t quite know how to become the one who needs something.
In that context, I can understand why an always-available conversation can feel comforting.
I would hesitate to dismiss that comfort simply because it comes through technology. Sometimes having somewhere to put your thoughts can help you get through a difficult evening. Sometimes writing something out helps you understand what you’re actually feeling. Sometimes practicing a conversation with AI makes it possible to eventually have the real one.
The concern, for me, begins when the practice conversation becomes a replacement for the real one.
The Relationships We Don’t Have to Risk Anything In
There is an unavoidable risk involved in being known by another person.
When I tell you something vulnerable, I don’t completely control what happens next. You may comfort me, but you may also misunderstand me. You may have your own feelings about what I’ve said. You may need time. You might challenge me. You may even tell me something I don’t particularly want to hear.
That can be uncomfortable.
But it is also part of intimacy.
Real relationships require us to encounter another person’s separateness. The person we love is not simply an extension of us. They have their own needs, limits, moods, histories and perspectives. Healthy relationships aren’t built by having someone respond perfectly every time. They’re built partly through what happens when they don’t: clarification, repair, compromise, apology, curiosity, frustration, forgiveness and trying again.
Technology can simulate aspects of a supportive conversation, but it cannot fully replicate the experience of being in a reciprocal human relationship.
It cannot genuinely need anything from you. It doesn’t have a childhood that collides with yours during an argument. It doesn’t have a nervous system that becomes activated when yours does. It doesn’t have to decide to forgive you after you’ve hurt it. It doesn’t experience the vulnerability of trusting you. And it cannot offer one of the most powerful experiences we have in relationships: another human being having the freedom not to choose us, and choosing to stay connected anyway.
Those things make relationships difficult.
They are also part of what makes relationships meaningful.
There is another reason I don’t believe AI can replace psychotherapy: in therapy, the relationship itself becomes part of the work.
How a person relates to their therapist can tell us something about how they experience relationships elsewhere. A client who is terrified that I’m disappointed in them may experience that fear in other important relationships. Someone who finds it almost impossible to disagree with me may have learned long ago that disagreement threatens connection. Someone who tells me they don’t need anyone may eventually discover how uncomfortable it feels to need something from me.
Those moments aren’t distractions from therapy. Very often, they are therapy.
The therapy room gives us an opportunity to notice those patterns while they are actually happening between two people. We can talk about them, misunderstand each other, repair things, experience frustration, negotiate boundaries and discover that a relationship can survive discomfort.
And there is an enormous amount of information in a therapy session that never appears in a transcript. I notice someone’s expression. Their posture. Their tone. The way their voice changes. Whether they suddenly become very still. Whether they are crying while insisting they’re fine. Whether a silence feels reflective, frightened, angry or disconnected. I notice what happens in the room when I say something that lands and what happens when I get something wrong.
Of course, therapists don’t interpret every crossed arm or moment of silence as psychologically significant, and we can misread things too. But therapy gives us something fundamentally different from a text exchange: a living relationship in which verbal and nonverbal information, history, context and clinical judgment are continually interacting.
AI can generate language that sounds empathic. It can help someone reflect. It can sometimes ask a useful question.
But it isn’t sitting in the room with you.
And sometimes what changes us isn’t the perfect response. It’s the experience of another person staying with us while we struggle to find the words.
When Support Becomes Avoidance
Avoidance doesn’t always look like running away.
Sometimes avoidance looks remarkably productive. It can look like researching, analyzing, journaling, reading another article, listening to another podcast, learning another attachment term, or having another long conversation about why we feel the way we do.
We can become extraordinarily insightful about a problem without ever doing the uncomfortable thing the insight is pointing toward.
You can spend an hour analyzing why you’re afraid to tell your partner that you feel neglected and still not tell your partner you feel neglected. You can understand exactly why you struggle with boundaries and still not set one. You can explore your fear of rejection in extraordinary psychological detail while continuing to avoid situations in which rejection is actually possible.
AI can become part of that process too.
It can help us prepare for difficult conversations, which can be incredibly useful. But there comes a point when preparation has to become participation.
Sometimes the next psychologically meaningful step isn’t another insight.
It’s sending the message. Having the conversation. Sitting through the silence after you’ve said something vulnerable. Hearing an answer you didn’t script. Allowing someone to misunderstand you and trying to explain yourself again.
Those experiences are messier than talking to technology. They’re also where a great deal of relational growth happens.
So, Should We Stop Talking to AI About Our Feelings?
I don’t think so.
I don’t think the most useful response to new technology is to pretend that people aren’t going to use it, or to shame them for finding comfort in it. AI is already becoming part of how people think, work, learn and, increasingly, process emotional experiences. There are legitimate ways it can support reflection and mental-health care.
The question I find much more interesting is what role we allow it to occupy.
Perhaps AI can help us find the words, but another person still needs to hear them.
Perhaps it can help us recognize a pattern, but we still have to practice doing something differently.
Perhaps it can reassure us once, while we also learn to tolerate the uncertainty that remains.
Perhaps it can keep us company during a lonely moment without convincing us that we no longer need companionship.
And perhaps the most important thing we can occasionally ask ourselves isn’t, “Is it okay that I’m talking to AI about this?”
Maybe it’s: “Is there someone in my life I need to be talking to about this too?”
Because there is something technology can offer us that is extraordinarily compelling: responsiveness without much relational risk.
Human relationships offer something different.
They give us the experience of being known by someone who is separate from us. They ask us to tolerate misunderstanding, uncertainty, disappointment and repair. They require us to discover that we can survive another person’s reaction. And, when those relationships are safe enough, they allow us to experience something no perfectly worded response can completely reproduce: being seen by another human being and discovering that we can still belong.
Maybe the goal isn’t to choose between technology and human connection.
Maybe it’s to make sure that one doesn’t quietly teach us how to live without the other.
Comments