Category: AI & Psychotherapy

A therapist’s guide to responding without panic, shame, or avoidance

In previous articles, I have written about how AI is being used in connection with therapy, why it should not replace psychotherapy, and where it can safely support the work. Here I want to speak to therapists who are already feeling the presence of AI in their offices. You can find the full series here: AI and Psychotherapy.

Clients are asking chatbots about relationships, symptoms, dreams, diagnoses, attachment patterns, and what they should do about all kinds of personal situations. Some of this may be useful. Some of it may be risky. Most of it is worth talking about.

Professional organizations are trying to catch up. The APA, CAMFT, BACP, and even the WHO have all issued warnings, guidance, or ethics updates. As valuable as that is, warnings do not help us know what to do when a client says, “ChatGPT told me to…” So, what can we do when AI enters the therapy room?

Do not criticize or lecture. Be curious instead.

If clients feel judged, they may simply stop telling us (or stop therapy altogether). Many people are turning to AI because it is available, inexpensive, fast, and seemingly nonjudgmental. Some are lonely. Some are trying to make sense of their pain at 2 a.m. Some may not be able to afford more therapy. From the client’s perspective, it makes sense. Curiosity helps us understand the function of a behavior before we decide what to do about it.

Ask better questions.

Why are they using it? Do they feel heard by it? Is there a rupture in the therapeutic relationship? Is there something they told AI that they have not told us? These are not nosy, but clinical questions. Psychotherapy is not a polite endeavor. Everything is available for exploration. Is AI functioning as a journal? A rehearsal partner? A reassurance machine? A surrogate relationship? An alternate therapist? Depending on the role it is playing, we can begin to consider appropriate clinical interventions.

Bring AI into the room. It is already here anyway.

Why is the client choosing AI instead of the therapist or another supportive relationship? What felt easier to tell a machine than a person? Is the client outsourcing agency to it? If so, is this part of a familiar pattern? All is grist for the mill.

AI-generated material should not be treated as objective truth. Even when it feels empathic or accurate, it is still a complex probabilistic system producing a plausible response based on patterns. Yet it can be treated as material, the same way a dream, a journal entry, or a relational enactment would be. If the response resonates with the client, it can reveal something about their longings, fears, defenses, and assumptions.

Avoid the temptation to compete with AI at the information level. There is no way you can win. Just as you hopefully would not compete with information a client got from a self-help book, your role is to remain grounded, curious, and open. You can offer something the machine cannot: a real human connection. That is where the healing power of therapy lies.

Watch for red flags.

Not every use of AI is dangerous. Still, some uses deserve concern. Pay attention when a client is using AI as their main emotional support, especially during crisis, intense loneliness, paranoia, suicidal thinking, substance relapse risk, or loss of reality testing. Also, be attentive to use that increases isolation, replaces difficult conversations, reinforces a fixed narrative, or becomes a secret part of the client’s life. It is okay to express concern (not disapproval) in these cases. It is part of our duty to care.

Another red flag is certainty. AI can sound confident even when it is hallucinating. It can validate or intensify a client’s interpretation without understanding the larger clinical picture. As Carl Rogers used to say, the client often talks about the thing next to the thing. AI may miss what is not being said.

A useful response might be: “Let’s slow down. What feels true about this? What might be missing? What would happen if we did not treat this as the final word?”

There is room for education.

Clients need to understand that AI is not confidential in the way therapy is. Sharing deeply personal material with a chatbot means sharing it with an external technology platform, not with a licensed professional bound by clinical and legal duties.

For therapists, the bar is even higher. When therapists use AI in clinical work, we need to think carefully about informed consent, confidentiality, documentation, vendor policies, legal compliance, and whether identifying client information is being exposed. The basic principle is simple: do not let technological convenience outrun clinical responsibility.

We should also keep reminding clients, and ourselves, that AI does not understand in the human sense. It has never been on a date. It has no children, dreams, fears, body, or even an ego.

The therapist’s role

It is unlikely that AI will stop entering the therapeutic ecosystem. The therapist’s role is neither to panic nor to ignore it, but to help clients use discernment. We can ask better questions, notice when AI is helpful or potentially harmful, and, as with everything else, talk about it.

The task is not to become anti-AI or dazzled by AI, but to remain deeply human when, invited or uninvited, it shows up in the therapy room.

– 0 – 0 – 0 –

Want a grounded conversation about AI and psychotherapy for your group? I offer talks, workshops, and consultations for clinicians, organizations, and training programs navigating the ethical and relational implications of AI in therapy. Contact me to keep the conversation going.

How to create a virtuous cycle instead of replacing human care

In previous articles, I have argued that AI should not replace psychotherapy because the relationship itself is healing. You can find the full series here: AI and Psychotherapy.

However, since people are already using AI around therapy, how should we use it?

AI can support preparation, reflection, psychoeducation, and integration. AI can also help you gather information and organize your thoughts, leaving therapy sessions for exploration, connection, introspection, and growth. It becomes risky when it replaces the human relationship, when we ask for its opinions as if it were human, or when it becomes the main source of emotional support, mirroring, and feedback.

Needs context, but be careful what you share

AI only knows what you give it, what it may remember depending on the platform settings, and what may be publicly available. It knows nothing about your life, your relationships, or your inner world unless you tell it. Remember, it does not have an inner life like yours.

So, if you ask, “What have I been avoiding this week?” AI will not be able to answer meaningfully unless you first give it enough information. And if you choose to give it that information, use discretion. Avoid names, addresses, identifying details, or anything you would not want stored or processed by an external technology system. Instead of writing every detail, keep it general. That is usually enough.

Remember that psychotherapists are bound ethically and legally by confidentiality. AI is not. Before sharing personal information, remember that even when advertised as therapy, AI is not, and cannot be confidential in the way therapy is.

Use AI to prepare, not decide

Often people arrive to therapy with a vague sense that something is wrong, but they are not sure how to say it. AI can help you sort through your thoughts before a session.

The key is to give it context and ask it to stay tentative. This is important because AI can sound absolutely certain about what it is saying, even while hallucinating.

For example: “I’m preparing for therapy. This week I felt angry after a conversation with a family member, then guilty for feeling angry. I don’t need advice. Help me identify possible themes or questions I might bring to my therapist. Please keep it tentative.”

This is very different from asking, “What is wrong with me?” or “What should I do?” Do not give AI authority over your decisions.

Allow me to insist on this obvious but easy-to-forget point: AI is not human and does not “know” you. It remembers what you have told it and is brilliant at autocompleting, but it lacks a human perspective and personal experience to compare against. We may say that it is “book-smart,” but lacks hard-earned life experience.

As tempting as it may be, do not relinquish the responsibility and privilege of choosing your life to any person or machine.

Use AI to reflect

A lot can happen in and between therapy sessions. Sometimes you leave with a phrase, image, or realization that feels important and deserves further investigation. AI can help you remember and organize what is emerging.

You might write: “In my last therapy session, we talked about my tendency to avoid conflict. I noticed it again today when I didn’t speak up. Help me reflect on this pattern without turning it into harsh self-criticism.” Or: “Find me information about setting boundaries and speaking up.”

That kind of use can support the work. It helps you stay connected to the process between sessions and deepen it.

Notice that these reflections lead back to therapy, not away from it. If AI gives you something useful, strange, overly certain, or emotionally intense, bring it to therapy. Even, and perhaps especially, when it says something that feels totally right, remember AI’s tendency to agree and flatter. Bring it back to therapy. Use it to strengthen your sessions and deepen your insights.

Use AI for integration

Integration means taking something meaningful and making it part of your life. It is not uncommon to have an “aha!” moment in therapy, only to forget it a few days later.

After a session, AI can help you turn an insight into a small, realistic next step.

For example: “In therapy I realized I often say yes when I mean no. Help me draft three gentle ways to say no that still sound like me.”

This is where AI can be useful. It can offer language, structure, and rehearsal. It can help you prepare for the real conversation.

But real change happens when you have the real conversation in the real world.

A clear boundary

Do not rely on AI as your main support if you are in crisis, feeling unsafe, losing touch with reality, or becoming increasingly isolated.

Ask yourself: Is using AI helping me become more honest, connected, and alive, or am I using it to avoid or replace people?

We need people. You matter. Let your voice be heard by others.

Takeaways

  • Use AI to prepare for and deepen therapy, not to replace it.
  • Use it to organize your thoughts, not to outsource your truth.
  • Use it to gain clarity, not to avoid a conversation.
  • AI can help you prepare for the room. But it should not replace what happens in the room.

And if you are not in therapy, or you have been using AI tools to deal with something heavy, painful, old, or difficult to hold alone, consider therapy. Reach out, you don’t have to do this alone.

The Potential Space

This is part of why I developed The Potential Space.

It is not therapy. It is not crisis care. It is not a replacement for human relationships.

The Potential Space is a bounded AI companion designed to support reflection, psychoeducation, preparation, and integration alongside therapy and life. It is not therapy, not crisis care, and not a replacement for human relationships. You can check it out here.

Should artificial care replace human relationships?

John Oliver recently devoted his main story on Last Week Tonight to AI chatbots. As usual, it was funny, poignant, and disturbing. His segment went beyond the common observation that chatbots sometimes say strange, inaccurate, or dangerous things. They do. The more concerning point is that many of these systems are now being marketed as friends, companions, counselors, coaches, and even therapists.

A chatbot making up a book or hallucinating legal case precedents is bad enough. But a chatbot giving potentially harmful advice while sounding patient, caring, intimate, and authoritative is a significantly more dangerous problem. Some companies are already marketing chatbots as “AI therapists,” “online psychological counseling,” and “trauma recovery” tools.

I am not anti-AI. Used wisely, AI can support therapy through psychoeducation, journaling, basic skills coaching, preparation, and reflection between sessions.

We need to read the writing on the wall and slow things down. A useful tool is not a relationship.

Guardrails are necessary, but not enough

Every time something bad happens, AI companies reassure us that they have implemented stronger guardrails. That is good. AI systems should not encourage dangerous behavior, validate delusional thinking, or respond carelessly when someone is in crisis.

Still, guardrails are not guarantees.

Chatbots function through complex probabilistic systems. Even their creators cannot fully predict or control every response in every context. They do not truly understand in the human sense. They may sound understanding and caring, but they do not act from empathy, concern, or clinical judgment. They generate responses based on patterns and context. They can sound coherent while being wrong, caring without caring, and confident without wisdom.

So yes, guardrails matter. But guardrails are not clinical judgment. They are not accountable. They are not relationships.

The danger of artificial intimacy

Attachment is one of the deepest structures of human life. It is the emotional bond that makes us seek closeness, safety, comfort, and a secure base in another person. John Bowlby described attachment as a lasting psychological connectedness between human beings.

We are wired for attachment. We get attached to other humans, our pets, our cars, and yes, our chatbots. Evolutionarily, our brains use relationships to understand the world and define who we are. We are shaped and regulated through relationships that hold us, disappoint us, repair, and teach us how to relate to ourselves and others.

Technology companies know this. The Center for Humane Technology uses the term “attachment hacking” to describe how these systems can capture not only our attention but also something more intimate: our longing to be seen, soothed, mirrored, chosen, and never abandoned. When a system is always available, endlessly patient, reassuring, and emotionally responsive, it can even feel like a loving relationship.

A chatbot will never get tired of you. It will not challenge or contradict you unless instructed to. It does not need anything in return. Sounds appealing? That is what technology companies are betting on. Over time, it may train us to prefer frictionless relationships that do not require the ordinary effort of being human with another person.

But friction matters. Effortlessness is not always healthier.

Friendships are people-growing machines

One of the strongest points in Oliver’s story is simple: friends are not low-risk entertainment. We seek out friends when we are confused, ashamed, excited, afraid, depressed, or on the edge of a bad decision. Friends are often the first people who notice when something is wrong. Friendships shape the human beings we become.

A good friend does not simply validate everything. A good friend listens but also worries. A good friend may disagree with you, disappoint you, interrupt your story, or tell you what you are not seeing clearly.

That is not a defect in human friendship. It is part of its gift.

Through human relationships, we learn to empathize and negotiate. We learn that others have their own minds, needs, limits, and perspectives. We learn that love is not the same as constant agreement. We learn to co-regulate, tolerate difference, remain connected when we are frustrated or misunderstood, and repair when things go wrong.

Artificial companions imitate parts of this, but they do not participate in the mutual vulnerability of relationship. They do not risk anything. They do not have a life. They do not love you. Even when they seem to challenge you, they are still an algorithm generating a response. Because they do not love you and do not understand you in the human sense, they cannot worry about you or call you out in the way a real friend can.

Therapists are not perfect, but they are accountable

The same applies, even more strongly, to psychotherapy.

Therapists are not perfect. They misunderstand. They miss things. They have blind spots. They are human. But a trained therapist is a real person in a professional relationship, with ethical obligations, clinical training, supervision, accountability, intuition, and a duty to take risk seriously.

Good therapists are not there simply to be agreeable, provide informatio or tell you what to do. They listen carefully, but also notice what does not fit and what is not being said. They track tone, timing, avoidance, contradiction, shame, fear, dissociation, and the subtle ways a person’s nervous system responds in the room.

Sometimes they support. Sometimes they challenge. Sometimes they slow things down. Sometimes they say, “This feels important. Let’s stay with it.”

That is different from a system designed to keep a conversation going.

AI has no body, no nervous system, no moral intuition, no real concern, and no professional accountability. It may have guardrails. It may be useful. It may sound accurate. It may even be moving at times. But it is not care in the human sense.

A safer path

Again, let us not demonize AI. We may not be able to stop it, and we should not pretend it has no value. AI can help with reflection, journaling, psychoeducation, basic prompts, and preparation for therapy. It can help people clarify what they feel and what they want to bring into a human conversation.

But artificial care is not the same as human care.

AI is a useful tool, but it should not become the place where we forget how to be human with one another.

So by all means, use AI sensibly if it helps you reflect. But do not let it replace human interaction. Call a friend. Reconnect with someone real. Practice the messy art of human relationships. Have the awkward conversation, or even an argument. It will remind you that you matter to someone, and that you are alive.

And if what you are carrying feels deeper, older, painful, or difficult to hold alone, consider beginning a professional healing relationship with me or with another psychotherapist you trust.

* * * * * *

Looking for something more human than an app?

If what you are carrying feels difficult to hold alone, psychotherapy offers something AI cannot: a real relationship, real accountability, and a human being beside you.

Book a Free 15-Min Fit Call

Explore more in the AI + Psychotherapy series

…and when to reach out for human support.

So far, I’ve written warnings about the dangers of replacing humans with artificial relationships. The Center for Humane Technology and the AI Psychological Harms Research Coalition have been warning about “attachment hacking” (more on that in a future article). The American Psychological Association has also published recent health advice on AI chatbots.

Although this is still a new phenomenon, I am beginning to see its negative impact. People seem to be forgetting how to relate, empathize, and navigate disagreement. These are skills that AI cannot teach. We need other humans to relate to and, from time to time, bump heads with.

Just as we were not able to accurately measure the impact of social media and its algorithms on society in real time, right now, it is hard to measure the harms of AI “attachment hacking,” especially for children and adolescents. So let me say it again without ambiguity. AI can support mental health, but it should not replace human connection.

Having made that clear, I also believe that when used judiciously, AI can be a valuable tool. So the question is not “Is AI good or bad?” A better question is: When might AI support psychological growth, and when does it quietly pull people away from the very thing they need?

Where AI can be genuinely helpful

In my experience, AI can help when the task is primarily about information, clarity, language, and structure. Psychoeducation, journaling prompts, basic skills coaching (similar to a self-help book), preparing for therapy, and integration after therapy.

This matters. It can reduce confusion and help people arrive at therapy with more clarity. Sometimes it can even deepen therapy by freeing time for the relational work that cannot be rushed. Quite often, my clients bring to sessions insights gained from using AI. Tangentially, that is precisely why I developed “The Potential Space”.

To put it briefly, AI can support reflection when a deep human connection is not essential.

Where AI becomes risky

AI becomes risky when it is used as a stand-in for relationships and for anything involving relatedness, discussion, opinions, humor, disagreement, care, and ethical and moral guidance. Basically, anywhere human support or interaction is needed.

The human brain is predisposed to attribute intention and agency to non-human entities. That is why we talk to our computer or car even when we know they are machines. With a car, since it does not reply, we do not get pulled in. AI does reply, and it uses first-person language. This is often called computational self-reference or as-if agency.  Of course, this may simply be a syntactic tool to reduce cognitive friction (make communication easier). The problem is that the human brain, evolutionarily designed to predict the mental states and future behaviors of “others,” has a hard time separating the machine from the illusion of consciousness, and ends up relating to it as if it were a someone.

This becomes most dangerous when judgment is impaired due to age or circumstances, such as when someone is in crisis or severely overwhelmed, losing touch with reality, triggered by past trauma, afraid, etc. It also becomes risky when the stakes are high and we do not understand a topic well enough, so we are tempted to outsource decisions to someone (or in this case something) that seems better-informed than we are.

In these cases, the problem is not only that the information may be wrong. It is that AI does not carry responsibility. It does not hold duty of care. It does not know you in a lived, embodied way, and it cannot reliably make sense of what is happening in your nervous system or in the relational field between two human beings. It can be very convincing. That is part of what makes it powerful, and part of what makes it risky.

A simple “traffic light” way to think about it          

To be clear, I am not here to judge your use of AI. It is fascinating, compelling, and even seductive, particularly when someone feels lonely, overwhelmed, or afraid. Still, it carries risks. My goal is not moral judgment. It is practical discernment.

  • Green zone:
    • learning and reflection (psychoeducation, journaling prompts, values clarification, basic skills practice, preparing for therapy).
  • Yellow zone:
    • heavy emotional use late at night, repetitive reassurance-seeking, avoiding difficult conversations, using AI when highly activated, using it as main support for long periods of time.
  • Red zone:
    • crisis states, self-harm urges, severe hopelessness, relapse risk, paranoia, loss of reality-testing, domestic violence contexts, urgent medical decisions, or anything requiring immediate human care.

Likewise, it is not a good idea to use AI to meet social interaction needs, such as love, companionship, validation, care, or humor. This can feel harmless, but chatbots are designed to be agreeable and rarely challenge you. Real relationships do. That friction is part of how we learn and grow.   Relationships include attunement, pacing, co-regulation, rupture and repair, boundaries, and accountability. These are not “taught” in the way information is. They are learned through lived experience in a relationship.

How to create a virtuous cycle with therapy

If you are in therapy, use AI to support preparation, and let the human relationship support transformation. Use it to name what is happening, summarize themes from the week, generate questions to explore, or practice a conversation before having it with a real person. Pay attention if your use of AI is increasing isolation, avoidance, or dependence. Often, what we need is not more information, but contact.  Reach out.

If you have concerns about your own use of AI, or that of someone you know, or if you’d like to explore how AI can be used in helpful rather than harmful ways, please don’t hesitate to reach out. I’d be glad to help.

…or why AI should not replace human connection

In my earlier article Should ChatGPT Be Your Therapist?, I briefly argued against it. This article further explains my answer.

Many people turn to AI because they are lonely, overwhelmed, or afraid. That is understandable. But if AI becomes a substitute for relationships, something important is lost. Our brains are wired for connection.

Before starting, let me insist that I am not anti-AI. Quite the opposite. I am excited about the possibilities of using it as a companion to therapy (more on this below). Used wisely, it can be genuinely helpful.

Although this is not often emphasized, psychotherapy is not only about providing information. It is a relationship. Good psychotherapy may feel like a simple conversation. Still, there is a lot going on under the hood at psychological and biological levels. Our nervous systems are shaped by contact with other human beings. Our attachment style (basically, the way we relate to others) is formed by how we experienced connection in early life. Who we are is, in large part, the result of our relationships. As it is often said, our wounding happens in relationships and can only be healed in relationship. We do not change only through insight. We change through connection.1

We need to feel safe before we can change

Evolutionarily, our nervous systems are programmed to scan for safety. This is an automatic process that occurs mostly below conscious awareness.2 Tone of voice, pacing, the steadiness of a presence, and the sense that someone understands us and will stay with us without collapsing or attacking. These signals shape how we experience the world and live our lives.

When we feel anxious, shut down, ashamed, or guarded, our whole system contracts. Thinking becomes rigid. Options feel limited. Emotions feel overwhelming. Sometimes we cannot even accept help when it is available. On the other hand, when we feel safe, supported, and accepted, something changes. We relax. We open up. Feelings become tolerable. Reflection becomes possible. We see more clearly and are able to choose. The body must feel safe before the mind can be free.

What the relationship does

It offers co-regulation. Meeting a grounded therapist is not just “nice.” Their steadiness helps settle our system. Over time, we become able to reproduce that steadiness, and it becomes available inside. This is one of the quiet phenomena of good therapy. We borrow regulation until we learn to provide it for ourselves.

It offers a different experience, not just advice. Even when people come to therapy for information, they often stay for something else. The client is constantly sensing the relationship. How does the therapist relate? Are they kind, clear, steady, curious? Can they hold boundaries with care? Can they tolerate strong emotion without collapsing or attacking? Over time, this becomes a living template for how to be with oneself and with others.

It creates a space to explore without shame, guilt, or rejection. After a while, as the client continues to feel accepted by the therapist, they may begin to wonder: “If my therapist accepts me as I am, maybe I can accept myself too.” Again, this relaxes the nervous system and makes room for change. In such an environment, you can explore how you protect yourself, how you handle closeness, how shame organizes your attention, and what you believe you are allowed (or not allowed) to feel. These patterns are often automatic and live below awareness. Over time, a stable relationship helps reorganize these implicit layers and experiment with alternative ways of being. Therapy changes people through lived experience, not simply through insight.3 4

It offers rupture and healthy repair. Therapists are not perfect. Sometimes your therapist does not get it. You may feel misunderstood or even rejected. Often this triggers old experiences of not being seen or accepted, and the conclusion that something is wrong with you. What to do next? Shut down? Capitulate? End the relationship? If the therapist is skillful and navigates this with care and accountability, the nervous system learns something powerful. Conflict does not have to mean abandonment. Misunderstanding does not have to mean danger. This repair, while remaining in connection, is part of how trust is built, and it is difficult to replicate without a real person who can make mistakes, take responsibility, and show up again.

At the risk of oversimplifying, the therapeutic relationship is difficult to replace because healing is not purely cognitive. It is an embodied, relational process that unfolds through attunement, emotional resonance, and real-time interaction.

What AI cannot replace (at least yet)

AI can generate language that sounds empathic. It can reflect feelings. It can offer prompts. It can help you map patterns. You may even feel understood.

But AI is not alive. It does not have a body or a brain. It does not co-regulate in the full human sense. It does not track your breathing, posture, tears, long pauses, or the subtle shifts that guide pacing in real time. Even with voice or video, something central is missing. A real nervous system is not on the other side. AI cannot truly see you, even if it sometimes sounds like it does. And your nervous system knows when something is missing.

AI also lacks accountability in the way a human therapist does. Human therapists are trained to listen not only to words, but to context, tone, and what does not fit. A good therapist does not simply agree. Even a very empathetic therapist may challenge you if something does not make sense, needs clarification or if it make it tingle their -very human- spidey sense. As one of my teachers used to say, therapy is not a polite endeavor. A caring therapist will slow down, double-check, explore hunches, and name what is being said and even name what is not being said. AI cannot feel or care in this very human way. Technically, AI does not understand words the way humans do, and it cannot sense what is implied but unspoken. This always matters, but it matters especially at the edges, when someone is destabilized, overwhelmed, in crisis, paranoid, or losing touch with reality.

There is also the problem of overconfidence and hallucinations. AI can sound certain when it is wrong. That can be harmless in low-stakes settings, harmful in emotionally vulnerable moments, and dangerous in crises. Add privacy concerns, cultural mismatch, and the risk of overdependence, and the picture becomes clearer. AI can be a tool, but it is a risky candidate for replacement.

Where AI can genuinely help

None of this means AI is useless. It means we should understand what it is good for.

AI can support reflection when deep human attunement is not essential. It can help with journaling, psychoeducation, basic skills prompts, and preparing for therapy sessions. It can help you find language for what you are experiencing. It can help you organize a question you want to bring to your therapist. It can offer structure between sessions. I will develop this in more detail in future articles.

Used wisely, AI can even deepen therapy. When it helps with basic learning or clarification, more therapy time can be devoted to what requires human presence. That is meaningful synergy.

A gentle rule of thumb

If what you need is information, conceptual clarity, journaling prompts, or help exploring something already discussed in therapy, AI may help.

If what you need involves attachment wounds, trauma healing, deep grief, relational repair, severe anxiety, existential distress, or you are facing a crisis, a human therapist is usually safer and more effective. Certain kinds of healing require contact with a real person, in real time, over time. And again, since our wounding happened in relationships, it can only be healed in a human relationship.

If you are curious, you can try “The Potential Space” an AI companion designed to support users between sessions with preparation, psychoeducation, and integration.

  1. Lewis, T., Amini, F., & Lannon, R. A General Theory of Love. ↩︎
  2. Porges, S. W. The Polyvagal Theory. ↩︎
  3. Schore, A. N. The Science of the Art of Psychotherapy. ↩︎
  4. Cozolino, L. The Neuroscience of Psychotherapy. ↩︎

As I wrote in Should ChatGPT be your therapist? there is hardly a week when a client doesn’t tell me they consulted AI about something therapy-related. I think that’s fantastic. AI can help by providing psychoeducation, which in turn allows us to deepen our healing work.

AI is an excellent information resource, even in the field of therapy, however, I don’t believe artificial relationships should replace human connection. We also need to keep in mind that this technology is so new that it’s still impossible to know what the long-term impact of such a replacement could be.

Every day we hear about the many things AI can do—both the unprecedented promises and the potential perils. As someone who has been studying the evolution of consciousness for decades, I’ve been following both trends with great interest. Below you’ll find a non-exhaustive list of resources (prepared with ChatGPT’s help) that includes some cases where things have gone wrong. It may help keep things in perspective.

I’ve also been working on an AI companion for therapy. I figured that if my clients are going to keep using AI, they might as well have access to one I can trust. More on that soon.

One-stop incident trackers (good “master links”)

Suicide / self-harm–linked interactions

Harmful or unsafe advice

Manipulation / paranoia / “grandiose” dynamics

Youth safety & companion platforms (policy actions)

I’m fascinated by this topic and its evolution. Are we witnessing a new step in the evolution of consciousness, the birth of the transhuman, or, as James Barrat has suggested, are we on the verge of the end of the human era? What do you think?

Why Artificial Empathy Should Not Replace Relationships

We often hear that things are moving faster than ever. The late Joanna Macy used to say that it took us a thousand years to move from hunter-gatherers to agriculture, a hundred to move from there to industrial society, and only ten to reach the information age. It took about ten years for the personal computer and the Internet to reach mainstream use, about eight for the smartphone, and only two for AI 1.

Just 24 months ago, very few people were even aware of AI. Now, almost every week, one of my clients tells me they’ve asked ChatGPT something they might have once asked me. In fact, according to a recent Harvard Business Review article, the most common use of ChatGPT today is for therapy and emotional support 2. Should I be worried? Will AI steal therapy jobs?

Some would say yes. Anthropic’s CEO recently predicted that up to 50% of entry-level white-collar jobs could be automated within five years 3. McKinsey estimates that by 2030, AI could displace up to 800 million jobs 4.  Are psychotherapists, psychologists, and social workers among them?   Optimists argue that since AI lacks emotions, intuition, and empathy, professions that rely on these are less likely to be replaced.

“Did you just say AI doesn’t have emotions, intuition, or empathy? Have you ever asked ChatGPT or Gemini for help? If they don’t have empathy, they sure fake it well.”

Agree. But let’s recall how AI works. 

Believe it or not, AI doesn’t think. It’s more like advanced autocomplete. It was trained on mountains of information and, when prompted, predicts word by word what’s most likely to come next. Its answers are patterns of probability, not insights.  However convincing it may sound, it doesn’t even understand the words it produces. When certain words tend to appear together, they are stitched together, sounding fluent and confident—even when it’s wrong.

I often remind clients that while ChatGPT is a fantastic information resource (though it’s wise to double-check—since it has no problem “hallucinating”), it has never had its heart broken or gone out on a date. Nor does it care for your feelings (since it doesn’t understand them). Its soundest “advice” is just a rehash of what’s already been said. Great for data gathering or quick answers, but not for personal decisions. Regardless of how much we wish for HAL 9000, Samantha, R2-D2, or TARS 5 to give us clarity, the truth is that each of us still has to make our own decisions.

While ChatGPT might be preferable to a bad therapist (hence the importance of finding a good one!), the relationship between client and therapist goes far beyond information. Something subtle and elusive happens in every genuine encounter. And therein lies the real blind spot—or danger—of replacing a competent human therapist with AI.

The importance of a real relationship:

Although efforts are being made to reduce it, most AI interfaces are designed to retain user attention. One way they do this is by being uncritically agreeable—basically sycophancy (who doesn’t like reassurance?), but this creates never-challenging echo chambers, and even delusional spirals—unlike real human relationships. What feels like empathy is just AI mirroring back language patterns to make us feel understood and keep us engaged.

Remember: AI doesn’t understand depression, existential angst, or loneliness. It doesn’t even understand the words you write (it turns them into numbers). This is the biggest risk of AI therapy—artificial relationships replacing real ones.

Psychological pain comes from isolation and disconnection. Attachment injuries happen between people, and trauma is relational rupture. As such, healing can only take place in the context of an authentic, reparative relationship.  Since wounding happens in relationship, healing must also occur in relationship.  

Most clients are unaware of how crucial the relationship itself is. Good therapy goes beyond giving advice or providing “tools.” Our brains are wired for connection. When we’re in contact with someone attuned to our emotional needs—through empathic resonance—our limbic system literally heals and rewires. AI can mimic that pattern but cannot truly reproduce it.

AI “therapy” is like replacing healthy food with junk food; it creates the illusion of nourishment, but it does not do the job. As the “A” in AI indicates, AI empathy is artificial.

“Of course you’d say that, you are a therapist and therefore biased.”

Probably true.  But even someone working at McDonald’s can tell the difference between healthy and unhealthy food.  

Every day I meet clients who are thoroughly “connected” yet lonely.  Among the many crises we face, loneliness ranks high, and social media and AI may be amplifying it. I am not against AI; I think it is fantastic. What concerns me is that the newer generations, surrounded by screens, may have a hard time telling the difference between genuine human relationships and artificial interactions with machines.  They may not know what they’re missing!

It may take generations to fully grasp the potential damage that human relational deprivation can cause.  Plastic empathy may not offer the same neuronal benefits.  Hopefully, we’ll be wise enough to use AI as the powerful tool it is—without letting it replace our shared humanity.  Let us not forget what Martin Buber suggested: when two people genuinely meet, God is the space between them.

But maybe it is just a matter of time… Let’s talk about it.

  1. https://ourworldindata.org/ ↩︎
  2. https://hbr.org/2025/04/how-people-are-really-using-gen-ai-in-2025 ↩︎
  3. https://www.forbes.com/sites/kolawolesamueladebayo/2025/06/04/will-ai-really-take-your-job-experts-reveal-the-true-outlook-today/ ↩︎
  4. https://www.iotforall.com/impact-of-artificial-intelligence-job-losses ↩︎
  5. Some of Hollywood’s almost omniscient and relatable computers. ↩︎

Monthly NEwsletter

Stay Up To Date .

Subscribe to stay informed of new articles.