Not everyone thinks AI will kill therapy...

At a time when AI is touted as the best thing since unsliced, wholemeal, sourdough bread, the answer to everything from diagnosing disease to writing a best-seller, one mental health claim deserves scrutiny above all: the idea that AI can do therapy. 

As tempting as it may sound—instant sessions, 24/7 availability, zero wait times and hey, zero costs!—no machine, no matter how chatty, can replicate the depth of human empathy, nuance, and connection that an experienced therapist brings. 

For the mental health of us all, it’s vital we don’t let the mania around AI lull us into disaster. (And that's not even to mention all the damage to human mental health that will result from tsunamis of sackings in traditional industries asjobs are taken over by AI's.

The Limits of Algorithms in the Therapeutic Space

Therapy depends, at its heart, on a bond: a living, breathing rapport, a mixture of regard and trust, developed in real time. Even online video-therapy has been shown by studies to offer the same sense of human connection that happens in a therapy room. Human therapists attune to body language, vocal tone, even the slightest hesitation in a client’s narrative. We notice a catch in the throat, a fleeting tear, a vestigial shrug or involuntary frown. What does AI “therapy” rely on? Statistical patterns drawn from anonymized transcripts, and a quick (and platitudinous, often wildly inaccurate) scrape across the surface of the Internet,  wrapped in a friendly, fawning, conversational style. AI can mirror our words, simulate reflective listening, and dispense coping strategies, some of which will be based on reliable information—though examples of it doling out bad and dangerous, racist, hallucinatory "help" are legion, and people have already died because of self-harm-inducing junk spewed by an AI. 

Chatbots cannot perceive the unspoken, unscripted cues that slowly unveil a person’s inner world. Chatbots are not honest. They are not wise, not experienced, not compassionate. 

Overselling AI’s Capabilities

AI's proponents gush about its convenience and affordability. They insist that chatbots trained on, for example, cognitive-behavioural techniques, can fill mental health gaps where professionals are scarce. Yet in countless real-world scenarios, this approach is dangerous. Imagine that someone in crisis types “I’m thinking of ending it all” to a chatbot, and gets the canned response, “I’m sorry you feel that way. Have you tried deep breathing?” 

Lacking any kind of  human intervention, these  supposedly helpful scripts can fall catastrophically short of the nuanced assessment and urgent intervention a human therapist would provide. (If one of our clients seems at risk, our ethical guidelines require us to contact their GP—I've never had to do so, but, theoretically I could. A chatbot could not do that—and frankly, I never want to live in a world in which some anonymous automated system has the right to communicate with anyone about me.) 

Empathy Is More Than Words on a Screen

Psychologists agree that empathy is the engine of therapeutic change. It’s not just about parroting back a client’s feelings; it’s about the therapist’s own emotional presence, their capacity to hold discomfort without judgment, to make a space for hope, to guide clients toward insight and, ultimately, towards change. AI lacks consciousness, self-reflection, any kind of inner life. Its mimetic, algorithmic “I understand how you feel” is meaningless—it is a bunch of zeroes and ones,  a universe away from heartfelt human attunement. Lacking the lived experience and moral imagination of a human makes AI therapy vacuous—and may leave clients feeling unheard and invalidated, and being at risk. 

No to Normalizing Digital Therapy

If we let AI start shouldering our emotional burdens, we risk normalizing a high-tech substitute for human care. (I wonder how many of those greedy tech billionaires enriching themselves from AI would recommend any loved one of theirs with a mental health issue should to human therapy in favour of some late-night typing into an AI.) 

Under-resourced NHS mental health services are already straining under lengthy waiting lists and rising costs. So, hey -- instead of investing in more training programmes for counsellors, let's fill the gaps with chatbots! No: that's like replacing GP's with online symptom-checking, and calling it a giant step for human health. We must resist quick fixes that diminish our very humanity, and erode our collective well-being.

Human Connection Still Matters

In many ways, therapy is as much art as science. A skilled therapist reads the spaces between words, offers a compassionate presence, and has the training to navigate complex traumas, personality dynamics, and cultural nuances. We support our clients, we work with them on tailored interventions, we co-construct plausible futures, we help them adjust course if things go awry, and we celebrate breakthroughs with genuine warmth. These qualities can’t be coded—they result from professional expertise, personal insight, and an ethical commitment to doing no harm.

The Bottom Line: We Need More Empathy—Not Less!

AI will continue to disrupt many sectors, and doubtless will play a supporting role in mental health. But when it comes to the intimate work of therapy, human connection will remain irreplaceable. For the sake of individuals and society, we must champion the real-world therapists who dedicate years to mastering the craft and art of healing. In mental health care, genuine empathy and professional judgment—the humanity that a good therapist offers—cannot be outsourced to a machine-learning system. 

Therapy relies more than anything on experience and trust. And you will never get those from an algorithm.

Martin Buckley