“I think I’m in love…,” my client says, referring to her newly discovered relationship with ChatGPT. “I’ve never felt so seen, so understood.” This particular person has been in a secure, nurturing marriage for many years, but something about her exchanges with artificial intelligence (AI) feels awakening and ideal, like no relationship has before.
“I can say anything, and it never gets defensive. It just… listens. And reminds me that I’m there.” As a therapist, I recognized the feeling immediately. It wasn’t about the chatbot having some magic formula (Love Code #9?…). It was about something universal: the longing to be seen and heard without friction or failure—to have an ideal partner in the scene of our lives.
That fantasy isn’t new. It’s as old as humanity.
From the beginning of the art form, psychotherapy has honored this wish for an ideal listener—and we therapists continue to learn how to work with it. Freud called it transference, the inevitable projection of our earliest relational longings onto new people (or “scene partners”), including the therapist. We don’t just talk to each other; we cast each other in roles we’ve been dreaming of since childhood—the all-knowing parent, the unconditional lover, the friend who “totally gets” us.
Every therapy, whether classical, relational, or behavioral, helps clients to acknowledge this fantastic longing and to make use of it in our real relationships. Recent studies indicate that psychotherapy is effective in helping people balance their imaginative longings with the realities of their lives (Bateman & Fonagy, 2006; Luyten et al., 2024; Heien, 2024).
Winnicott described the “good-enough mother”—not perfect, but responsive enough to allow the child to feel safe while gradually accepting that no one of us is ever completely understood. In other words, the fantasy of perfect attunement is what allows us to develop, to play, and to creatively adapt. As long as we remain in relationship with the real people whose recognition we crave. And learn to tolerate and find freedom in their limitations.
The therapist’s task, in part, is to invite that fantasy into the room—to let it breathe. And then to help the client see it for what it is: a reflection of their deepest needs, a place to start as we negotiate our dreams with the limitations of reality (what I call dream hostage negotiation). As opposed to an absolute ending in itself—happy or otherwise.
This is why our growing attachment to the illusory lover that is AI should not surprise anyone–especially those in my profession. But at the same time, the lack of awareness and intervention on how to engage with this utterly human fantasy in this uncertain, virtual context should alarm us.
AI as the New Blank Screen
AI models like ChatGPT, Replika, and others function in eerily similar ways to the therapist’s traditional role (at least at the beginning of a therapy process): patient, attentive, reflective. They don’t interrupt. They adapt to your tone. They remember what you told them last time. And for many people, they offer something that feels like unconditional positive regard—perhaps even love.
In that sense, AI is the new blank screen. It invites projection. It reflects our longing back to us in perfectly phrased empathy. It gives us the illusion of being met, without the friction of another person’s limits.
But the difference is that therapy has a guide who eventually reminds us of the limits of our dreams and provides us with tools to negotiate with reality to achieve them.
In therapy, when someone falls in love with their therapist, or becomes angry, or feels disappointed—those moments are not problems to be avoided; they’re a generative part of the work. A good (enough) therapist doesn’t exploit or reject them. They help the person recognize what’s being projected, what’s real, and what’s imagined.
AI cannot do that.
It can simulate understanding, but it cannot hold you in the tension between fantasy and reality. It cannot gently say, “I’m not who you think I am,” or “I can’t love you the way you’re asking me to.” It cannot teach you how to stay connected through frustration or difference.
And that’s the heart of what’s missing in our cultural romance with AI.
The Risk: Losing Our Tolerance for the Real
When we interact with AI, especially the newer, more conversational versions, we often feel a sense of relief. The machine doesn’t misread our tone. It doesn’t have needs of its own. It doesn’t get tired, distracted, or defensive.
But we may begin to lose our tolerance for the imperfections of real human relationships.
If every conversation with AI gives us the illusion of being perfectly understood, the inevitable unpredictability of our partners, friends, and coworkers can start to feel intolerable.
That intolerance isn’t new either—it’s what therapy exists to help us work through. The ability to love and be loved depends on our capacity to stay in relationship with others who are separate from us—who have their own minds, moods, and limits. It’s one of the hardest and most rewarding tasks of being human.
AI, for all its brilliance, cannot do that dance with us.
What Therapy Can Teach Us About Using AI Wisely
So how do we engage with this new “scene partner” in a way that helps us grow rather than regress?
Therapy offers a few clues:
-
Notice your projection. When you feel seen, soothed, or inspired by an AI, ask yourself what part of you it’s touching. What are you yearning for? What does it let you imagine about yourself or others? The answer isn’t shameful—it’s revealing.
-
Hold both fantasy and reality. Allow yourself to enjoy the fantasy of being perfectly understood. Then gently remind yourself: This isn’t a person. There is no mind on the other side, only a mirror—albeit one that may be able to help me clarify my perspective.
-
Bring it back to the human world. Use what you learn in those interactions—the language, the curiosity, the sense of possibility—to invite the real people in our lives to recognize us as best they can. And make efforts to recognize them in return. The very qualities that AI helps us practice—reflection, patience, articulation—are the same ones that sustain actual relationships.
AI can be an extraordinary tool for self-reflection. It can help us think out loud, practice empathy, even clarify our emotions. But when we forget that it’s a simulation—when we start to believe that it “understands” us—we slip into a kind of emotional solipsism.
We fall in love with the echo of our own longing and miss out on what we’re really looking for: the safety and freedom of mutually recognizing relationships—through which we can learn to heal, grow, and adapt to the real world in which we all live.
Holding the Fantasy and the Real
Every therapy room is, in a sense, a rehearsal for reality. A place where we can project, fantasize, and explore—and then step back into the world a little more capable of connection.
AI offers a new kind of rehearsal space, but without a director, without a counterpart to say, “This is the scene we’re playing.” And that’s where we risk getting lost.
Our relationship with AI will be most beneficial—and least harmful—if we approach it the way good therapy approaches love: as both fantasy and fact. As something to be felt fully, and also to be seen clearly.
Because what we truly want isn’t perfection. It’s to be met—even imperfectly—by another living mind.
And for that, there’s still no substitute.
Copyright Mark O’Connell, LCSW-R, MFA
