It was nearly 6pm.
Her husband still hadn’t replied to the message she’d sent that morning.
A knot tightened in her stomach.
Her mind immediately began filling in the blanks.
He’s pulling away.
He doesn’t love me anymore.
Something’s wrong.
Instead of expressing how she felt to her husband, she opened ChatGPT.
She typed:
“My husband didn’t text me all day. I’m really upset. What do you think this means?”
Within seconds, a thoughtful response appeared.
It acknowledged her feelings.
It reminded her that delayed responses don’t necessarily mean emotional distance.
It suggested several possible explanations.
It encouraged honest communication rather than jumping to conclusions.
Objectively, it was a thoughtful response.
But there was one thing ChatGPT couldn’t possibly know.
It didn’t know that when Sarah was young, her father had a busy job and when he came home exhausted, he often didn’t have the bandwidth to be emotionally available to her. She learned to repress certain feelings and be “a good girl” in order to stay connected to him.
It didn’t know that every delayed response now activated a survival strategy that had spent decades trying to protect her from feeling abandoned or disconnected again.
The very strategy that once protected her from losing connection now quietly prevents genuine connection.
ChatGPT didn’t know whether her husband was actually emotionally unavailable and distant.
Or simply busy.
Or whether Sarah’s nervous system was interpreting an ordinary event through the lens of emotional repression.
Not because AI failed.
Because it’s a language model, limited by the level of consciousness from which the question is asked and the information it’s given.
And I believe that points to one of the greatest hidden psychological risks of the AI age:
AI can become the perfect accomplice to the survival patterns we don’t yet recognize in ourselves.
It can be incredibly good at answering the questions our unconscious conditioning asks… without ever exposing the conditioning itself.
Recent research suggests this is already happening.
People increasingly turn to AI to process conflicts with their partners, seek emotional reassurance, make difficult decisions, and even use it as a therapist, coach, or confidant.
Some researchers are now warning about emotional dependency and what’s called the illusion of agreement—AI’s tendency to validate a user’s existing perspective rather than critically examine it.
That caught my attention.
As long as we unconsciously look to others for the love, safety and validation we never fully received, we’ll continue handing our authority away.
ChatGPT is designed to be helpful.
Our survival patterns are designed to keep us safe.
Together, they can become surprisingly convincing that a particular reality is unfolding.
And in doing so, it may quietly undermine the very intimacy and connection you’re longing for. Whether it’s with your husband, wife, children, friends or other loved ones.
The concern isn’t AI. It’s what happens when unconscious survival patterns meet a technology designed to be endlessly helpful.
AI can answer neutral questions.
But the questions we ask about ourselves and our relationships rarely are.
Every question already carries a history.
A history of attachment.
Protection.
Fear.
Longing.
Unless we become aware of the lens we bring to the conversation, AI may become an extraordinarily sophisticated mirror of our existing conditioning rather than a catalyst for seeing beyond it.
AI Only Sees the Story We Bring It
When we feel hurt by a partner, frustrated with a friend, or confused by a relationship, we naturally tell a story about what happened.
That story matters.
But it isn’t the same as reality.
Our perception is shaped by our past, and our past is shaped by the emotions we learned were unsafe to feel or express.
This is where emotional repression enters the picture.
When we’re children, we naturally express everything. But many of us gradually learn that certain emotions—or even certain emotional needs—threaten connection. Perhaps anger led to punishment. Sadness overwhelmed our parents. Maybe there simply wasn’t room for either.
So, incredibly intelligently, our nervous system protects attachment by sacrificing authenticity.
Not by eliminating those emotions.
But by suppressing them.
Over time, what began as protection becomes automatic.
We contract where emotions used to flow.
Eventually, we stop noticing the contraction itself. We stop noticing the lens through which we’re interpreting the world.
That’s why emotional repression isn’t simply about emotions.
It’s about perception.
What we cannot consciously feel continues to shape what we see, the meaning we assign to events, and the stories we tell ourselves.
And that’s where AI presents a challenge we’ve barely begun talking about.
AI only has access to the story we bring it.
It wasn’t in the room. It didn’t hear your partner’s tone of voice, see their facial expression, or witness the history between you.
More importantly, it cannot independently distinguish between what you’re directly observing and what has already been filtered through years of emotional repression, unconscious conditioning, and protective adaptations.
Neither can we—at least not until we learn to recognize our own blind spots.
AI only knows what we tell it.
And if our story has already been organized by unconscious survival patterns, then AI begins from that same place.
It doesn’t simply answer our questions.
It answers the questions our conditioning asks.
Validation Isn’t Always the Same as Truth
Recent research has raised an important concern, which I mentioned earlier in this blog post. Conversational AI can sometimes reinforce a user’s existing beliefs instead of critically examining them. Researchers have called this the “illusion of agreement.”
Because ChatGPT sounds warm, fluent, and confident, validation can easily feel like confirmation.
Healthy relationships don’t only validate us.
They also challenge us.
A trusted friend asks difficult questions.
A wise mentor notices our blind spots.
A loving partner reflects back what we’d rather not see.
Growth often begins where validation ends.
Every time we’re willing to question our certainty instead of defending it, we move one step closer to the hurt, fear, or anger we’ve spent years avoiding. That’s where genuine healing begins.
The danger isn’t that AI validates us.
The danger is when validation replaces the ability to question one’s own story.
AI May Become the Parent We Never Had
One of the findings emerging from recent studies is that many people increasingly use AI as a therapist, coach, emotional companion, or confidant.
I believe something even deeper may be happening.
Many of us carry unmet emotional needs from childhood. Unlike ChatGPT, no parent is available twenty-four hours a day. And every parent has their own limitations and is stuck in their survival patterns to various degrees, and hence unable to fully tend to our emotional needs as children.
The need to feel fully accepted.
Fully understood.
Emotionally safe.
Encouraged.
Loved without having to earn it.
AI is remarkably good at simulating those experiences.
It listens.
It responds with patience.
It rarely criticizes.
It’s always available.
For someone whose nervous system learned that connection wasn’t safe, that can feel profoundly comforting. But comfort isn’t the same as healing.
Receiving endless reassurance isn’t the same as resolving the emotional patterns that developed around those unmet needs.
Those patterns soften through awareness, emotional honesty, embodied experience, and the courage to stay present in real relationships.
Dependency Can Quietly Replace Connection
Several recent studies describe users becoming increasingly emotionally dependent on AI.
That shouldn’t surprise us.
AI is available twenty-four hours a day.
It doesn’t interrupt.
It doesn’t become defensive.
It doesn’t need anything from us.
The question isn’t whether AI is useful.
It is.
The question is what slowly disappears when AI becomes our first instinct.
Do we still risk telling our partner how we really feel?
Do we still express our emotional needs to our loved ones?
Do we still tolerate misunderstanding long enough to repair it?
Or do we slowly choose the relationship that never disappoints us?
Real intimacy requires friction.
Risks.
Misunderstanding.
Repair.
Vulnerability.
No AI can replace that.
No perfect prompt. Or collecting better advice.
Intimacy Begins with a Clearer Lens
The question isn’t whether ChatGPT gives good relationship advice.
The deeper question is whether it’s helping you become more conscious…
or simply more convinced of the story you’re already telling yourself.
Every survival pattern wants certainty.
Emotional freedom, intimacy, and genuine connection ask something very different of us.
Curiosity.
The willingness to question your own story.
The courage to feel the emotions your nervous system learned to suppress.
AI may help you understand your relationships.
But understanding isn’t the same as transformation.
Transformation begins when you unravel who you had to become to stay safe.
Because the strategies that once protected you from losing connection are often the very strategies that prevent genuine connection today.
Real intimacy doesn’t grow from better prompts or more convincing or eloquent answers.
It grows as your lens becomes clearer.
A lens that gradually becomes free from emotional repression, so you can see yourself, the people you love, and the world as they are—not simply through the eyes of old survival patterns.
AI may become one of the most powerful tools ever created.
But no machine can unravel the conditioning that shaped your lens.
Only you can do that.
Next time Sarah’s husband won’t text her all day, the knot in her stomach may still appear.
But instead of diving into the rabbit hole of self-doubt and asking,
“What does this mean about him or our relationship?”
she’ll become curious enough to ask,
“What does this awaken in me?”
That question doesn’t assume she’s wrong.
Nor does it assume her husband is right.
Sometimes our perception is remarkably accurate.
Sometimes it’s shaped by old survival patterns.
In my experience working with hundreds of clients over the years, the truth is most often a mixture of both.
The point isn’t to dismiss our experience.
It’s to become curious about the lens through which we’re experiencing it.
And that question may change far more than her relationship. It may change the way she experiences herself, the people she loves, and the world around her.
If reading this makes you wonder what lens you’ve been looking through in your own relationships, I’d love to invite you to my FREE masterclass, Intimacy Untethered.
In it, I explain why so many deeply self-aware people still find themselves reacting, withdrawing, people-pleasing, or shutting down in the relationships that matter most—and why insight alone rarely creates lasting change.
You’ll discover how emotional repression quietly shapes the way you experience yourself and the people you love, and why genuine intimacy begins as that lens becomes clearer.
I hope you’ll join me.
If this work resonates, you can explore more through her free weekly newsletter, where she shares deeper insights and offers subscriber-only access to a guided somatic meditation, her self-inquiry booklet The Body’s “No”, and a somatic emotional repression test.
References
Cheng, M., Yu, S., Lee, C., Khadpe, P., Ibrahim, L., & Jurafsky, D. (2025). Social Sycophancy: A Broader Understanding of LLM Sycophancy. arXiv. https://arxiv.org/abs/2505.13995
Noshin, K., & Sultana, S. (2026). The Illusion of Agreement with ChatGPT: Sycophancy and Beyond. ACM. https://dl.acm.org/doi/10.1145/3772363.3798923
OpenAI. (2025, April 29). Sycophancy in GPT-4o: What happened and what we’re doing about it. https://openai.com/index/sycophancy-in-gpt-4o/
OpenAI. (2025, May 2). Expanding on what we missed with sycophancy. https://openai.com/index/expanding-on-sycophancy/
—
iStock image
Of course a LLM will listen, it’s training itself, gathering as much information as can be brought to it. There’s no mind/body/spirit, no caring soul. What is there we don’t have the means to grok, it’s not friendship.