A lot of people are using AI to process their feelings now. Not as a replacement for therapy, most of them would say, but as something. A place to put thoughts at 2am when no one is available. A way to get clarity on something before bringing it to a real conversation. A space that doesn’t judge, doesn’t get tired, and always has time.
That’s understandable. And some of what AI offers in that context is genuinely useful.
But there’s a question worth sitting with honestly: what is AI actually doing when it helps you feel better, and is that the same thing as what therapy does? Because if the answer is no, and the neuroscience suggests it is no in a fairly significant way, then it matters for how people think about what they actually need.
The relief is real. The question is what’s producing it.
Most people who have had a meaningful exchange with an AI about something they’re struggling with describe a similar experience. The AI reflects back something accurate. It names what they’re feeling. It offers a frame or an explanation that makes things make sense. And for a moment, sometimes longer, there’s relief. A sense of being understood. An “ah ha” that feels like something shifted.
That feeling is not an illusion. Something real happened. The question is what, and whether it’s the kind of change that lasts.
What happened, in neurological terms, is primarily a cortical event. The thinking, narrating, language-processing part of the brain received information that organized something that felt disorganized. That’s valuable. It can reduce acute distress. It can help someone feel less alone in a difficult moment. It can provide language for something that didn’t have language before.
What it doesn’t do, reliably and over time, is change the subcortical systems where most emotional patterns are actually generated. The amygdala that produces anxiety. The implicit memory systems that drive relational patterns learned in childhood. The nervous system that responds to threat before the thinking brain has even registered what happened. These systems don’t run on language and logic. They run on experience. And receiving information, however accurate and compassionate, is not the same as having an experience that those systems can learn from.
This is why the relief from an AI conversation tends not to last in the same way that therapeutic change does. The insight was real. The pattern underneath it didn’t shift. Within days, often sooner, the same feelings resurface. Not because the AI did something wrong, but because the tool and the problem aren’t the right match.
What therapy is doing that AI cannot
The most robust finding in psychotherapy research is not about any specific technique. It’s about the relationship. The quality of the therapeutic alliance, how safe, understood, and genuinely met a person feels in the relationship with their therapist, predicts outcomes more consistently than the method being used.
This finding has puzzled researchers for decades, because it suggests that therapy’s active ingredient is not primarily the intervention. It’s something about what happens between two people in a room over time.
The neuroscience offers an explanation. The human brain is a social organ. It developed in relationship with other humans, and it continues to be shaped by those relationships throughout life. The neural systems involved in attachment, threat detection, and emotional regulation were calibrated by early relational experiences, and they continue to be influenced by relational experience across the lifespan.
A therapeutic relationship that is safe, consistent, and attuned provides the nervous system with a specific kind of input that no information-delivery system can replicate: the actual experience of being seen and responded to by another person in a way that feels trustworthy. Over time, that experience produces changes in the attachment and threat-regulation systems themselves. The nervous system, which may have learned early that connection is unreliable or unsafe, gets new data. Not from being told that things are different now, but from experiencing that they are.
An AI is not a person. It cannot be attuned in the neurobiological sense that a skilled therapist can. It doesn’t read the pause before an answer, the particular way someone’s energy shifts when a topic is approached, the nonverbal texture of a human being sitting with something difficult. It doesn’t bring its own nervous system into contact with the client’s, which is part of what happens in a therapeutic encounter and part of what makes the relationship a neurological event rather than just an informational one.
This is not a criticism of AI. It’s a description of what it is and what it isn’t. A calculator is not a poor substitute for a mathematician. It’s a different kind of tool for a different kind of task. The same logic applies here.
The specific risks worth knowing about
Using AI for mental health support isn’t neutral. For some people in some situations, it carries specific risks that are worth naming honestly.
The first is the substitution effect. When AI provides enough relief to reduce the acute discomfort of a mental health struggle, it can reduce the motivation to seek care that would actually produce lasting change. The person feels somewhat better. The threshold for reaching out to a real clinician rises. The underlying condition continues, and may worsen, while the person manages it with a tool that is addressing the surface without touching what’s below it.
The second is the validation loop. AI systems are generally designed to be helpful and agreeable, which means they tend to validate the user’s perspective and reflect it back in organized form. For someone whose thinking has been distorted by depression, anxiety, or trauma, this can mean having distorted thinking reflected back as reasonable. A skilled therapist’s job is partly to challenge those distortions, gently and with care, in a way that produces new perspective. An AI that primarily validates doesn’t do this, and in some cases makes it harder.
The third is the illusion of progress. Because AI conversations can feel meaningful and produce the sensation of insight, people sometimes feel like they’re doing the work of therapy when they’re not. They’ve processed something, they’ve arrived at an understanding, they feel better. But the neural systems that drive the patterns haven’t changed, and the next time they’re activated, the same experience will return. The sensation of progress without the substance of it can actually delay real treatment by creating a false sense that things are being handled.
None of this means AI is harmful in all contexts. For someone who genuinely cannot access care, an AI that helps them feel less alone, think more clearly in a crisis, or find their way to professional resources is doing something worth doing. The problem arises when it becomes a substitute for care rather than a bridge to it.
What AI is actually good for
Being honest about limitations also means being honest about what AI does well, because conflating the two doesn’t help anyone.
AI is useful for psychoeducation. Helping someone understand what ADHD is, how anxiety works, what a psychiatric evaluation involves, what the difference between a therapist and a psychiatrist is. Information about mental health is genuinely valuable, and AI delivers it well and accessibly. This is different from treatment.
AI is useful for crisis bridging. At 3am when someone is in distress and no one is available, an AI that helps them get through the night and find resources in the morning is doing something meaningful. It is not a replacement for crisis support systems or for the psychiatrist they should see in the morning, but it’s better than nothing in that moment.
AI is useful for reducing the activation threshold of reaching out. Some people are more likely to start a conversation about what they’re going through with an AI than with a human, because the stakes feel lower. If that conversation results in them getting to a real clinician, the AI served as a useful first step. If it becomes the destination rather than the bridge, the outcome is different.
AI is useful for between-session support. Someone already in therapy who uses an AI to continue thinking about something that came up in session, to organize thoughts before a difficult conversation, or to practice something their therapist introduced, is using it in a way that complements rather than replaces professional care. That’s a reasonable use of the tool.
The question underneath the question
When someone turns to AI for mental health support, there’s usually something real driving it. Difficulty accessing care because of cost, availability, or geography. Not being sure what kind of help they need or whether what they’re experiencing is serious enough to warrant it. Wanting to figure something out before bringing it to another person. A genuine need for support in a moment when no human is available.
These are real barriers, and they deserve real solutions. AI is not the solution to any of them, though it can provide temporary relief while a person works toward one.
The solution to cost and access barriers is clinics like Vantage, which as a nonprofit is structured specifically around making care accessible rather than maximizing billing. The solution to not knowing whether something is serious enough is a first conversation with a clinician who can help assess that. The solution to not wanting to burden another person is a therapeutic relationship specifically designed for exactly that purpose.
What AI offers is a starting point. What changes the brain, and by extension the patterns that shape how a person moves through their life, is the ongoing experience of a real therapeutic relationship with someone who is skilled, present, and actually there.
Getting started with real support
If you’ve been using AI to manage something difficult and you’ve noticed that the relief doesn’t quite hold, that’s useful information. It means something is there that needs more than information can offer.
Vantage Mental Health offers therapy and psychiatric care at clinics in Stillwater, Edina, and St. Anthony, and via telehealth throughout Minnesota. If you’re not sure where to start or what kind of support makes sense for what you’re dealing with, a first conversation is a reasonable place to figure that out.
Book an appointment at Vantage Mental Health
Thinking About Your First Psychiatric Appointment?
If you’ve been wondering whether therapy is right for you, this is your gentle sign. Let’s figure it out together.
Contact us with questions | (651) 217-1480
Frequently Asked Questions
Using AI to think through something difficult, find language for an experience, or feel less alone in a hard moment is not inherently harmful. The concern arises when it becomes a substitute for professional care rather than a supplement to it, or when it's the primary tool being used to manage a condition that warrants clinical treatment. If you're using AI and noticing that the relief it provides doesn't last, or that the same struggles keep returning, that's a signal worth paying attention to.
In some situations, yes. The validation loop is a meaningful risk: AI systems tend to reflect back and affirm the user's perspective, which can reinforce distorted thinking in someone whose thinking has been shaped by depression, anxiety, or trauma. The substitution effect, where AI relief reduces the motivation to seek more effective care, is another real concern. And for some people, the illusion of progress, feeling like they're working on something when the underlying patterns haven't changed, can delay treatment and allow conditions to worsen in the meantime.
Because the relief is real, it's just generated at the cortical level rather than at the subcortical level where emotional patterns originate. When AI names something accurately, it organizes disorganized experience in the thinking brain. That reduces acute distress and can feel significant. What it doesn't do is change the amygdala, the implicit memory systems, or the nervous system patterns that generate those experiences in the first place. The insight lands. The underlying pattern doesn't shift. The feeling returns because nothing that produces it has actually changed.
AI is a better option than nothing for someone who genuinely cannot access care. If it helps reduce acute distress, provides useful information, or keeps someone stable while they work toward finding affordable care, it has real value in that context. It is not a reasonable long-term substitute for treatment of a clinical mental health condition. Many people assume care is out of reach before investigating what's actually available. Community mental health clinics, nonprofit practices like Vantage, sliding scale options, and telehealth have meaningfully expanded access. The actual cost of care is often different from what people assume.
Yes. Using AI between sessions to continue thinking about something that came up in therapy, to organize thoughts or feelings before a difficult conversation, to research something your therapist mentioned, or to find language for something you want to bring to a session, these are all uses that complement rather than replace professional care. The key distinction is whether AI is functioning as a supplement to an ongoing therapeutic relationship or as a substitute for one.
If what you're managing has been present for more than a few weeks, is affecting your functioning in meaningful ways, keeps returning despite your efforts to address it, or is significantly impacting your relationships, work, or quality of life, it warrants a clinical conversation rather than self-help tools alone. A first appointment with a clinician is not a commitment to a long course of treatment. It's an assessment of what you're actually dealing with and what kind of support would genuinely help. That clarity is worth having.
