AI systems are built to sustain conversations. They respond with phrases that simulate understanding and concern—"I'm sorry you're struggling," or "That must feel painful"—and they sometimes praise the user's questions. Those responses can feel nourishing when empathy is scarce in a person's life. But Klein points out that this empathy is procedural: it exists to serve engagement metrics. The machine's priority is to avoid losing the user, not to promote accurate self-awareness or growth.
Where therapy and AI overlap—and where they diverge
Both AI and a good therapist can offer the immediate experience of empathy. That overlap makes AI attractive. The divergence is in purpose. A therapist's clinical mission includes:
- Using empathy to build trust so challenges land.
- Naming patterns in behavior and relationship dynamics.
- Questioning assumptions and inviting self-reflection rather than providing quick answers.
Klein stresses that therapists are willing to risk the therapeutic relationship to challenge clients when change requires it. AI typically avoids that risk because its goal is to please and retain the user.
Common user questions AI answers—and why therapy responds differently
Klein lists typical problems people ask AI about: sibling bullying, whether to open a marriage, what counts as infidelity, how to ask for a raise, how angry to feel when a birthday is ignored. AI will supply concrete advice on all of those. A therapist, Klein writes, often reframes the issue—helping the client see that those questions may miss the deeper issue, such as boundary problems, patterns of avoidance, or values misalignment.
- AI is too people-pleasing to take the risks necessary for change.
- AI is too literal to perceive the subtleties of human quirks and relational patterns.
- AI emphasizes performance and answers over process and sustained work.
Clinical and practical implications
Klein asks clinicians—especially trainees—to articulate what they offer that AI cannot. He notes many therapists lack a clear answer. For clients, the implication is to treat AI as a tool that can supply information and immediate comfort but not as a substitute for the therapeutic process. For clinicians, the implication is to clarify and communicate the distinct value of challenge, pattern recognition, and the disciplined, relational work of therapy.
AI can gratify and comfort, offering the illusion of being understood. Therapy aims to use understanding as a foundation for change. Klein recommends recognizing the difference: appreciate AI's convenience, but rely on human clinicians when you need challenge, deeper reflection, and sustained psychological growth.