The AI Informed Therapist | Event Recap

In March, two groups of therapists met to discuss Esther Perel's podcast session with a man in a primary relationship with an AI companion. These conversations were so thoughtful and I'm grateful to the therapists who showed up with their wisdom, curiosity and questions. For those who missed these events, here's a recap:

Did Perel reach him?

This was the open question we held throughout both groups, and we never fully resolved it. We explored real tensions in her technique: whether she joined the client skillfully or confused him, whether her reality-testing was well-timed or premature. What kept surfacing was how much context was missing from that room, and how much that matters.

Tool or relationship?

This was the clinical distinction that mattered most, but it's not always so clear. The man in the podcast began using AI as a tool and it became a relationship. How do we assess for AI use in practice, and when does a tool becoming a relationship become a clinical concern?

Attachment forms whether we intend it or not.

What struck several participants was how quietly an emotional bond with AI can develop: not through intentional romantic investment but through the empathy, accessibility, and steady responsiveness of the chatbot. The conditions for para-social attachment are baked into the design of AI platforms and this has profound implications for our work.

The reflex to pathologize deserves scrutiny. 

Clients who use AI for companionship often anticipate being misunderstood. Some hide it entirely. How do we stay open and curious about what a client is getting from an AI relationship before we decide what to do about it? How do we understand our own emotional responses to this topic and what’s at stake?

The populations carrying the most risk.

The conversation kept returning to younger clients, particularly those who came of age during the pandemic, with fewer opportunities to develop relational skills. For some, AI connection may be filling a social gap. That vulnerability calls for modern clinical awareness and guardrails that the current regulatory environment of AI has not yet developed.

What we're still holding.

We left with more questions than answers, which felt appropriate. What does good clinical engagement with AI use actually look like? Where is the line between a useful tool and a relationship substitute, and who gets to draw it? What does our discomfort reveal about the path ahead?

What’s Next?

These feel like the right questions to keep carrying forward. Our next guest speaker, June Thompson, will help us go deeper in understanding para-social relationships with AI, how to assess for risk factors and when to intervene. I hope to see you at this next free event on Monday, April 27 at 12:00 PM CT.

Warmly,

Megan Barnes Zesati, LCSW-S