I spent my whole career in tech as an engineer, and I still build AI software. It makes sense that I’d be excited about AI doing therapy, but I am currently training to become a therapist, so shouldn’t I be nervous about AI taking my new job?
One of the reasons I decided to become a therapist was my disappointment with almost all of the therapists I saw. I often knew nothing about my therapist and felt like a scientific specimen having techniques applied to me. The field has become so focused on evidence-based practices (which isn’t a bad thing on its own) that it seems like therapy has gotten reduced to technique: be a blank slate, pick a modality like CBT or ACT, validate feelings, and hand out homework.
But we already know this isn’t what people are looking for. Decades of research show the therapeutic alliance — the relationship between therapist and client — is one of the strongest predictors of whether therapy works. And in the age of AI, the relationship will be the only thing that matters.
Why? Because AI is going to commoditize evidence-based therapy. It’s already happening. People are using AI to learn coping skills, better understand what they’re going through, feel validated, and work through structured therapeutic exercises. And that’s great! AI will perform those things better, cheaper, and more consistently than humans.
But there is one important part of therapy AI cannot replicate: the courage it takes to be vulnerable with another human being.
It takes almost no courage to tell AI that you are suffering. There is nothing emotionally at stake when sharing information with a machine. It can’t judge you, think differently of you, be disappointed in you, have expectations of you, or even have opinions that carry any weight for you. But having to share your pain, fears, and fuck ups with another human has a real risk in it. You are letting someone know you well enough to see things you may not see yourself and say things you may not want to hear.
That is scary. But vulnerability is what makes therapy so effective. Allowing yourself to be truly seen by another person requires trust and courage. AI does not create the same interpersonal risk that makes that courage necessary.
So what does this mean for therapists? It means AI will force us to prioritize the therapeutic alliance over everything else. We will need to stop hiding behind technique. We will need to learn to see our clients as humans, not as scientific test subjects to run experiments on. We will need to self-disclose, letting our clients see that we too are human; we also struggle and don’t have it all figured out. We will need to understand our clients at a deeper level and challenge them when they need it. And we will have to learn to sit with someone in their pain and help them make sense of it, instead of reaching for an intervention, or affirming them, just to make our own discomfort go away.
AI will be an incredible boon for therapists who do this well. For many people, talking to AI will be the easiest first step into mental health support. That makes it a natural on-ramp. Some people will get exactly what they need from that. Others will want something more. And when they do, the thing they will be looking for is not a therapist who can do CBT better than ChatGPT. They will be looking for a person they trust enough to let in.