The other day, something happened that genuinely made me stop and think about the future of therapy.
Like many people, I’ve found myself using AI from time to time to help organize my thoughts or look at relational dynamics from a fresh angle.
Understandably, AI is a loaded and deeply complicated topic right now, and I want to acknowledge that. My own clients have a wide range of reactions to it, from outright terror, anger, and disdain to celebration of the ways it has radically supported their lives. As a neurodivergent therapist who works with many neurodivergent (ND) clients, I have also seen how AI can be a genuinely transformative tool for navigating some of the disabilities and barriers that are often central to the ND experience.
At the same time, there are profound ethical and existential dilemmas surrounding its use. I will readily admit that I am still trying to make sense of many of these questions myself. I am sitting in the muddy tension of what AI may mean for our humanity, relationships, society, and the earth. I do not have a tidy answer or a definitive line drawn in the sand, but I still want to acknowledge the complexity rather than writing about AI as though that complexity does not exist.
On this specific day, I was sitting with a text exchange that had left me feeling a bit activated. I copied the conversation into AI, the therapist in me curious what kind of insight or feedback it would offer. And I have to admit something that some therapists may be reluctant to say out loud:
AI did a remarkable job at making sense of themes, analysis and offering insight.
In fact, there were parts of its analysis that I thought were every bit as thoughtful as what I might have come up with myself– and perhaps even better. It recognized patterns I hadn’t fully put words to yet, synthesized the interaction in a matter of seconds, and reflected back a theme that felt incredibly accurate. There was nothing particularly magical about what it was doing. It was simply very, very good at recognizing patterns across a large amount of information. Then it said something that landed squarely on one of my own attachment sore spots and almost immediately, I teared up. It got me… genuinely.
In that moment my partner happened to walk into the room, saw me welling up, wrapped me in a hug, and jokingly said:
“Well… I guess AI is about to take your job.”
We both laughed, but I’ve thought about that moment a lot since then because I actually think it illustrates something incredibly important about what therapy is, and perhaps more importantly, what therapy isn’t.
The insight that AI delivered wasn’t what made me tear up. The insight simply opened the door, and what happened next after the insight is what actually landed.
My nervous system didn’t actually didn’t need another explanation. In truth, I already knew the insight it reflected back. In that moment, I didn’t need AI to generate five more paragraphs or give me another beautifully organized summary of my family dynamics. In that moment, my body instinctively reached for something much older and much more fundamental:
It reached for another human being. It reached for my partner, for co-regulation, for human connection.
As an attachment therapist, I don’t think that’s a small point, but rather the point.
One of the things I’ve noticed over the past year is that more and more of my clients are walking into session having already used AI. They’ll often pull out their phones and say, “Can I read you what Claude said?” More often than not, I find myself genuinely impressed. Sure, there’s a validation vending machine element to AI. We know that AI is designed to hook you in and provide you with a certain amount of feel-good responses, lacking in the challenge and activation that comes with a real human therapist.
But on a cognitive level, most of the analyses are thoughtful, balanced, emotionally intelligent, and surprisingly nuanced. I rarely find myself disagreeing with the overarching conceptualization.
What’s fascinating, though, is that as clients read those responses aloud, sometimes like me, they well up. It can feel really powerful to feel understood, even by a pattern-recognition robot. When the insight and sense of “being seen” lands, all of a sudden they are no longer trying to understand themselves but trying to figure out what to do with the grief, shame, anger, loneliness, or fear that the insight has awakened.
That is where I believe my work actually begins.
Many of us grew up in families that taught us how to explain, suppress, outrun, or distract ourselves from emotions, but very few of us were taught how to actually move through them, especially in a relational context. We become remarkably savvy at understanding ourselves while simultaneously remaining overwhelmed by ourselves and stuck. We know exactly why we feel abandoned, why conflict is activating, why we overfunction in relationships, or why we shut down when someone gets too close to us. Yet despite all of that insight, we still find ourselves lying awake at two in the morning with a nervous system that doesn’t feel any different.
Insight, as valuable as it is, is not the same thing as integration.
Integration, especially when dealing with atttachment trauma or injury, requires co-regulation, attunement and the presence of a safe, secure attachment figure. We learn regulation and how to move through emotional experiences differently within the container of safe connection. And so as much as AI might give CBT and insight-oriented therapists a run for their money, they can’t well up with you in the room. They can’t look you in the eyes offering limbic resonance and a deep sense of being seen by a human. AI cannot offer you co-regulation, attunement, genuine connection.
And so, no: I’m not afraid that AI is going to take my job.
