AI
Why AI Chatbots Cannot Replace Human Psychotherapists
Simulated empathy may ease disclosure, but genuine therapeutic change demands real human connection over time.
Key takeaways
- AI chatbots can simulate empathetic responses and make it easier to disclose difficult feelings, but this is not equivalent to therapy.
- Genuine psychotherapy requires navigating discomfort, conflict, and emotional work that AI systems cannot replicate.
- Lasting psychological change depends on being known over time by a human professional, not just receiving momentary support or validation.
- The ease of disclosing to AI may actually undermine therapeutic effectiveness, as productive discomfort often drives real change.
- Organizations should view AI mental health tools as supplements or triage mechanisms rather than replacements for human therapists.
The limits of algorithmic empathy
Artificial intelligence tools can now generate responses that sound empathetic and validating. These capabilities make it easier for people to disclose feelings they might otherwise struggle to express. But according to psychologist Helen Marlo, writing in Psychology Today, this simulated understanding falls short of what genuine psychotherapy provides.
The distinction matters because psychological change requires more than supportive words. Real therapeutic work involves navigating discomfort, working through conflict, and engaging in emotional labor that AI systems fundamentally cannot replicate.
Why it matters
As mental health chatbots proliferate and companies market AI as a scalable solution to therapy access gaps, understanding what these tools can and cannot do becomes critical. Organizations investing in employee mental health benefits need to know whether algorithmic support constitutes adequate care—or whether it risks substituting convenience for effectiveness.
What therapy actually requires
Marlo identifies a core element missing from AI interactions: being genuinely known over time. Therapeutic relationships develop through sustained connection where a human professional comes to understand a client's patterns, contradictions, and growth trajectory.
This differs fundamentally from receiving support, advice, or affirmation in isolated exchanges. An AI can validate feelings in the moment, but it cannot hold the accumulated knowledge of who someone is across months or years of sessions. It cannot notice when current struggles echo past patterns, or recognize growth that the client themselves hasn't yet articulated.
The therapeutic process also depends on productive friction. Real change often emerges from moments of discomfort—when a therapist gently challenges a client's narrative, sits with difficult silence, or names patterns the client would prefer to avoid. These interventions require human judgment about timing, relationship strength, and individual readiness that algorithms cannot meaningfully assess.
The disclosure paradox
Marlo acknowledges that AI chatbots do lower barriers to initial disclosure. People may find it easier to type vulnerable thoughts to a bot than to say them aloud to another person. This reduced friction has value, particularly for those who might otherwise avoid seeking any support.
But ease of disclosure should not be confused with therapeutic efficacy. The very discomfort that makes human therapy harder is often what makes it work. Saying difficult things to another person, and having that person respond with both empathy and appropriate challenge, creates conditions for genuine psychological change.
Implementation considerations
For organizations evaluating mental health technology, these limitations suggest AI chatbots may serve best as triage tools or supplements to human care rather than replacements. They can help people articulate concerns, provide psychoeducation, or bridge gaps between therapy sessions—but not substitute for the relationship-based work that drives lasting change.
The analysis was first reported by Helen Marlo in Psychology Today, based on her clinical expertise as a psychologist at Notre Dame de Namur University.