How AI Is Making Me a Better Therapist, Not Replacing Me as One
There is a lot of conversation right now about whether AI is going to replace therapists.
I use AI almost every day, and strangely, using it has made me more convinced that it can’t.
Because one of the most valuable things AI does in my practice isn’t therapy.
It makes me examine how I do therapy.
And sometimes it calls me out.
Which, frankly, is exactly what I want it to do.
With appropriate client consent, privacy protections, and HIPAA-compliant technology, telehealth can give me something therapists historically haven't always had easy access to: the ability to go back and look at what actually happened in the room.
Not what I think happened.
Not the version I reconstructed while writing my note.
What actually happened.
And after a session, I can essentially take that clinical material and ask AI:
Okay. What did you see that I didn't?
I Don't Want AI to Tell Me I'm a Great Therapist
That would be completely useless.
I want it to challenge me.
What relational patterns were happening underneath everything everyone was talking about?
What did I catch?
What did I miss?
Where did I move too quickly?
Did I accidentally align too strongly with one person in this couple or family?
Was I explaining something the client actually needed to experience?
Did I rescue someone from discomfort too quickly?
Was there another way to formulate what I was seeing?
Where might my own countertransference have entered the room?
What should I come back to next week?
Basically, I don't need a robot cheerleader.
I need another surface for my brain to bounce against.
And sometimes AI gives me an observation that makes me think, Oh. That's interesting.
Sometimes it points out something I completely missed.
And sometimes I read what it says and think, Absolutely not.
That part is useful too.
Because disagreeing forces me to articulate why I disagree. What do I know about this client that changes the interpretation? What history matters here? What did I feel in the room that isn't captured in the words? What theoretical framework am I using that leads me somewhere different?
AI isn't making the clinical decision.
It's making me defend my clinical thinking.
There is a difference.
Apparently, I Have Patterns Too
This might be the most uncomfortable and useful part.
When you do this repeatedly, you stop just noticing patterns in your clients.
You start noticing patterns in yourself as a therapist.
Imagine that I bring multiple sessions to AI and keep getting some version of:
You identified the pattern accurately, but you got there before the client did.
Well.
That's no longer information about the client.
That's information about me.
Maybe I need to slow down.
Maybe my brain connected all the dots and got excited and I handed someone the answer instead of giving them enough room to discover it themselves.
And knowing something intellectually is not the same as experiencing it.
That's therapy 101.
It is also apparently something therapists occasionally need to be reminded of.
Or maybe I start noticing that I become particularly protective of the overfunctioning person in a family.
Why?
What am I responding to?
Am I seeing something accurately, or is something about that role pulling on me personally?
Maybe I keep doing a beautiful job explaining an emotion when what the client actually needs is for me to stop talking and let them sit inside it for 30 seconds.
That is a clinical growth edge.
And it's an incredibly specific one.
Continuing Education From My Actual Tuesday
I love training. I love books. I love learning new clinical approaches.
But there is something very different about attending a continuing education training, taking 14 pages of notes, putting them in a folder, and hoping that three months from now your brain magically retrieves page nine at exactly the right moment.
AI lets me learn from my actual work while the work is still fresh.
Something comes up in a session that I don't fully understand?
I can go deeper.
Show me the research on this.
What books should I read?
How might a structural family therapist conceptualize this compared with an IFS therapist?
What am I missing about this particular attachment pattern?
What does the literature say?
Is there an area here where I actually need more training?
That last question matters.
Because AI isn't always the destination. Sometimes it points me toward the destination.
A book.
A research article.
A consultation.
A training.
A supervisor.
An entirely different way of thinking about something I've been looking at the same way for years.
It Has Changed What Happens Between Sessions Too
One of the frustrating things about therapy is that we can have this incredible moment at 10:43 on a Tuesday morning where everything clicks.
And then the client goes back to work.
Someone needs to be picked up from soccer.
Dinner needs to happen.
A spouse says the exact thing that activates the exact pattern we just spent an hour discussing.
And by Sunday, that beautiful insight from Tuesday is somewhere in the same part of the brain where we store passwords and the reason we walked into Target.
AI helps me make therapy more tangible.
I can take something meaningful from our session and turn it into a reflection question.
A small experiment for the week.
A coping strategy to try.
Language to use when the same argument starts again.
A follow-up email.
A way of practicing the thing we talked about instead of just remembering that we talked about it.
But here's the really important distinction:
AI doesn't decide what gets sent. I do.
It might generate twenty ideas.
Being a therapist means knowing which one belongs with this particular human being.
Or deciding that none of them do.
Because I know the client.
I know their history. I know their relationships. I know their readiness. I know their sense of humor. I know what happens when they feel pushed. I know what we're working toward. I know when homework will feel supportive and when it will feel like one more thing they're failing to accomplish.
AI has information.
I have a relationship.
Those aren't interchangeable.
And Yes, There Are Very Real Concerns Here
I don't think we should talk about AI and mental health without talking about the responsibility that comes with it.
Privacy matters.
Informed consent matters.
HIPAA matters.
The technology being used matters.
Accuracy matters.
Bias matters.
Clinical judgment really, really matters.
AI can also be spectacularly confident while being completely wrong.
Anyone who has spent enough time using it knows this.
So no, I don't think therapists should be feeding protected health information into random AI tools, blindly accepting whatever an algorithm says, or outsourcing clinical decision-making because ChatGPT produced a very convincing paragraph.
AI should never become an unquestioned clinical authority.
Neither, frankly, should therapists.
The goal isn't to stop thinking because technology can think for us.
For me, the value has been almost the exact opposite.
Used well, AI forces me to think harder about my own thinking.
What am I missing?
What assumptions am I making?
What does the other person in this system experience?
Have I become overly attached to one formulation?
Is there another explanation?
Am I actually helping right now?
Or am I overfunctioning?
Those are questions good therapists should be asking themselves anyway.
AI has simply given me another place to ask them.
The Human Part Matters More, Not Less
The more I use AI, the less interested I become in the question of whether it can "do therapy."
Because therapy isn't just saying the psychologically correct thing at the psychologically correct moment.
AI can't sit across from my client and feel the room change.
It doesn't know what it means when this client suddenly stops making eye contact.
It doesn't know that the joke they just made is how they tell me we're getting too close to something.
It doesn't have years of relationship stored in the space between us.
It doesn't feel protective of someone and then have to wonder why.
It doesn't carry responsibility for that person's care.
It doesn't know them the way I do.
And they don't know it.
The relationship matters.
The human part matters.
If anything, technology has made me more intentional about protecting that part of the work.
Because if AI can help me research, organize, brainstorm, notice patterns, generate possibilities, and review my own work, then I have fewer excuses for being on autopilot when an actual human being is sitting across from me.
I can be more curious.
More prepared.
More reflective.
More willing to question myself.
More present.
AI isn't replacing me as a therapist.
But it can hold up a pretty damn useful mirror.
And becoming a better therapist requires being willing to look in it.
