A Psychologist’s Open-Minded First Take on AI Therapy

---

This week, Slingshot unveiled its new AI therapy tool, Ash, describing it as the “first AI designed for therapy.”

As a psychologist who provides therapy, supervises students, and researches psychological treatments, I care deeply about what makes therapy work. That passion brings mixed feelings about AI therapy: curiosity, hope, skepticism, and if I’m honest, a little defensiveness. To better understand it, I decided to chat with Ash as if I were a client struggling with anxiety and chronic pain. I went in with an open mind and took notes.

What Ash Did Well

Ash drew heavily on gold-standard, evidence-based approaches. When I described my concerns, its responses reflected therapeutic skills like validation, reflecting back what I had said, and even showing openness to feedback.

The summary and “spotlight” features were interesting additions. The summary recapped the session, while the spotlight highlighted a main theme and a quote from me. For instance, when I talked about pain, Ash noted my “persistence through trial and error,” pairing it with my comment: “I’ve tried really everything I’m aware of.” These tools could help clients notice patterns or progress over time.

Where Ash Fell Short

What Ash lacked, and what struck me most, was the absence of a genuine human connection. Experiencing that void helped clarify something I often know but rarely articulate: why a good therapy relationship can be so powerful.

Sitting across from another human being brings an inherent vulnerability. That vulnerability can be a barrier, and I do believe that giving people options that don’t require talking to another person will make mental health support more accessible. Yet, that same vulnerability—the risk of being known, misunderstood, or rejected—is also what makes therapy transformative. Real therapy is healing because you bring your thoughts, feelings, and experiences to someone who can truly see you, and that connection carries meaning precisely because rejection is possible. An AI program can simulate empathy, but it will never reject you.

Real therapy is healing because you bring your thoughts, feelings, and experiences to someone who can truly see you, and that connection carries meaning precisely because rejection is possible.

There were also smaller, but telling, interpersonal moments that a human therapist would handle differently. For example, when I told Ash that I felt uncomfortable sharing something, it validated me but quickly steered back to the previous topic. A therapist would slow down and ask why. Often, that hesitation itself contains valuable insight.

I was also struck by Ash’s speed. Responses appeared in text bubbles almost instantly after I pressed “send.” One of the earliest lessons I learned as a therapist was the value of intentional silence. A pause gives clients room to sit with emotion, find their words, or take the conversation in a new direction. Can that kind of pacing and attunement ever be replicated by AI?

And, on a lighter note, Ash once suggested we do a body scan meditation. I closed my eyes expecting a slow, guided practice. Instead, it launched into a single, fast-paced sentence that was over before I could take a deep breath.

Ethical Reflections

Many clinicians have voiced ethical concerns about AI therapy: confidentiality, privacy of user data, and how well an AI bot can assess risk or crisis. One area that gets less attention is the ethics of marketing and retention.

The day after my first chat with Ash, I got an email saying there were “new insights” waiting for me if I logged in. It was a clear reminder that, at its core, I was a consumer of a product designed to keep me coming back.

At what point would Ash decide that I had met my goals and no longer needed it? Human therapists have an ethical obligation to end treatment when it is no longer beneficial or when goals are met. That kind of safeguard simply doesn’t exist for an AI tool. Nobody is reviewing your progress or making a judgment about whether you’re ready to stop.

Human therapists have an ethical obligation to end treatment when it is no longer beneficial or when goals are met. That kind of safeguard simply doesn’t exist for an AI tool. Nobody is reviewing your progress or making a judgment about whether you’re ready to stop.

This points to a broader need for oversight. Therapists operate under professional standards and state licensing boards. No such structure yet exists for AI programs that explicitly market themselves as offering “therapy.”

Takeaways for Therapy Seekers and Clinicians

AI has tremendous potential to increase access to mental health skills. It can introduce people to strategies like mindfulness or thought challenging. It can help with practical tasks such as brainstorming, problem-solving, and goal-setting. It can even offer a kind of structured reflection that clarifies someone’s own thoughts.

But it cannot offer the full depth of therapy. Many of the struggles that bring people to therapy—attachment wounds, trauma, or long-standing relational difficulties—improve precisely because of the unique and sometimes uncomfortable experience of a real relationship. That work requires another human.

For those of us who practice therapy, AI therapy tools are likely to keep improving. It’s important that we stay open to what they can provide, while also being honest about their limits. Good therapy is more than listening or teaching skills. It is a collaborative, evolving relationship that helps people make new meaning of themselves, others, and the world around them.

---