I’m going to start with something that might surprise you: I’m actually a big fan of ChatGPT.
I use it myself. I appreciate having a tool that helps me organize my thoughts, consider different perspectives, and challenge my assumptions. I don’t want it simply agreeing with me. I want it asking questions and helping me recognize what I might be missing.
I’ve always been interested in the books, movies, podcasts, and other resources my clients use to make sense of their lives. Sometimes I’ll explore those resources myself. Sometimes I’ll challenge the conclusions they’ve drawn. And sometimes we just have a good laugh about something they’ve discovered.
One client even told me she’d customized her ChatGPT to sound like me! I had to laugh, but I appreciated that she felt comfortable telling me.
To me, that’s part of building a genuine therapeutic relationship. I want clients to know I’m interested in their world, not just what happens during our sessions.
AI is becoming another part of that world.
And if we respond to clients’ use of AI with judgment or dismissiveness, they may simply stop telling us about it.
We don’t have to endorse every AI-generated conclusion to welcome the conversation about it.
Because lately, I’ve noticed something concerning.
Clients are arriving with AI-generated explanations of their partners’ behavior, communication scripts, and sometimes what they believe is proof that their partner is the problem.
They may leave an AI conversation feeling more certain without necessarily becoming more accurate.
When AI Becomes the Authority on Your Partner
Imagine a woman who worries that her partner isn’t being completely honest. She describes their interactions to ChatGPT and eventually concludes that he’s engaging in trickle truthing.
She brings that conclusion into therapy, convinced ChatGPT has identified what’s happening.
Her concerns may be legitimate. There may be dishonesty that needs to be addressed.
But ChatGPT has only heard her account.
It has received her interpretation of events, filtered through her emotions, memories, and assumptions. It hasn’t heard her partner’s experience or explored what happens between them.
Yet its response may sound authoritative.
As therapists, we know that validating someone’s subjective reality isn’t the same as confirming every conclusion they’ve drawn from it.
Are clients asking AI to help them understand their relationships, or to confirm what they’ve already decided?
There’s an important difference.
The List Isn’t the Relationship
I see a similar problem when couples approach relationship advice like a checklist.
Consider a couple where one partner wants to feel more appreciated.
His partner takes that feedback seriously. She creates a list of appreciative statements, saves them in her phone, and reads them to him.
She’s trying. She’s remembering. She’s doing what she believes he asked her to do.
But he doesn’t feel particularly seen or valued. The appreciation feels like a task she’s completing rather than an emotional connection she’s offering.
He withdraws.
She becomes frustrated.
“But I’m giving you the appreciations! I even have them written down!”
And suddenly, they’re back in their familiar pattern.
He’s shutting down because the connection isn’t landing. She’s becoming defensive because her effort isn’t being recognized. Neither understands what’s happening for the other.
The list isn’t the relationship.
AI can provide wonderful examples of appreciative statements, softened startups, and repair attempts. But knowing the right words isn’t the same as knowing how to use them in a relationship.
Sometimes You Need a Therapist in the Room to Put All the Pieces Together
When I’m sitting with a couple, I’m not just listening to their words.
I’m watching the relationship happen.
I notice the missed eye contact, the shift in body language, the moment someone’s tone changes or a partner becomes quiet.
I’m paying attention to signs of flooding, bids for connection that aren’t received, the Four Horsemen showing up, and opportunities for repair that slip by unnoticed.
And I’m trying to understand how those pieces fit together.
That doesn’t mean I automatically know what someone’s body language means. It means I can get curious and ask.
“I noticed that when she read those words, you looked away and became quiet. What was happening for you?”
Then I might turn to his partner.
“When he got quiet, what did you tell yourself that meant?”
Now we’re doing something different.
We’re not debating whether she completed the appreciation assignment correctly. We’re helping them understand why her effort isn’t creating the connection either partner wants.
Maybe he needs to express his positive needs more clearly. Maybe she needs help understanding how her tone affects the message. Maybe both need support tolerating discomfort long enough to hear each other.
We can slow the interaction down, help them recognize defensiveness, practice a softened startup, and make room for repair.
This is where Gottman Method interventions become more than concepts couples have read about.
We’re not just teaching couples what healthy communication sounds like. We’re helping them recognize what gets in the way of experiencing it together.
Are We Analyzing Text Messages Instead of Having Conversations?
Another pattern I’m seeing involves clients uploading text exchanges to AI and asking what their partner really meant.
Was that message manipulative? Was it dismissive? Who was wrong?
But think about how much can be lost in a text.
A simple “Fine” might mean agreement, frustration, exhaustion, or an attempt to end the conversation.
AI can suggest possible interpretations. It can’t independently establish what the sender intended.
Sometimes my intervention isn’t to help a client analyze the message more carefully.
It’s to ask, “Would this conversation be better handled directly with your partner?”
Of course, context matters. Written communication can help with regulation, and direct conversations aren’t appropriate in every situation, particularly when safety is a concern.
But I want couples to consider whether they’re using technology to build understanding or avoid the vulnerability of talking with each other.
Helping Clients Use AI to Become More Curious
I don’t think we should tell clients to stop using ChatGPT.
Instead, let’s help them become better consumers of the information they’re receiving.
Consider the difference between asking:
“Here’s what my partner did. Explain why they’re manipulating me.”
And asking:
“Here’s how I experienced an interaction with my partner. What assumptions might I be making? What alternative explanations should I consider? How might my own defensiveness be contributing? What could I ask my partner to better understand their experience?”
The first question seeks confirmation.
The second invites curiosity.
Neither guarantees an accurate response. AI can still be overly agreeable, miss important context, or offer interpretations with more confidence than the information warrants.
But the questions we ask influence the kind of reflection we’re inviting.
As therapists, we can encourage clients to bring AI-generated insights into sessions as possibilities to explore rather than verdicts to defend. We should also remind them to consider privacy before sharing sensitive relationship information or a partner’s messages with an AI platform.
And when concerns involve abuse, coercion, or safety, clinical assessment must take priority over generic communication advice.
Being Right Isn’t the Same as Being Connected
I appreciate AI. I think it can be a valuable companion for reflection, preparation, and learning.
But relationships aren’t simply problems to solve with better information.
They’re living interactions between two people who need to feel heard, valued, and understood.
Someone can memorize a softened startup and still deliver it with criticism. They can practice appreciation and still miss their partner’s emotional needs. They can learn the language of repair without recognizing the moment a repair is needed.
Sometimes couples need someone to help them slow down, notice what’s happening between them, and put the pieces together.
AI can help people prepare for a better conversation. It cannot have that conversation with their partner for them.
Our goal isn’t to help clients become increasingly certain about what’s wrong with their partners.
It’s to help them become more curious about each other, more accountable for their own contributions, and more capable of creating the connection they’re longing for.
Because being right about your partner isn’t the same thing as being connected to them.
