The Conversation Therapists Should Be Having About AI
The Real Fear Isn't Replacement
The American Psychological Association ran a survey with 1,200 psychologists, and 77% reported that their patients have spoken to them about using AI to support their mental health needs. More than a third of the psychologists also said their patients are using AI as an additional mental health professional. This is something we really need to talk about.
I expected professional therapists to be worried about being replaced by AI. It seemed like a reasonable assumption, given how widespread the fear of AI replacing human jobs has become. I thought the survey would show that therapists felt their livelihoods were at risk, creating a natural survival instinct to reject AI altogether.
Interestingly, the survey found that fear of replacement was not what concerned psychologists most.
Instead, 97% of psychologists were concerned that chatbots used for mental health could inadvertently reinforce negative behaviors, validate delusional beliefs, or contribute to self-harm. The survey identified several specific concerns, but they all pointed to the same underlying fear: that chatbots would not know how to respond appropriately and could ultimately cause harm.
These aren't the wrong fears. They're exactly why people should be talking to AI that's purpose-built for emotional support.
Not All AI Is Created Equal
As a therapist who works closely on the training, safety, and user experience of Abby, our AI for mental health, I'm not surprised that colleagues who haven't had the opportunity to look behind the scenes of AI would be cautious about clients turning to it for support. Before becoming deeply involved behind the scenes at Abby, I likely would have shared many of those same concerns.
The difference is that once I pulled back the mysterious curtain, I found a team of smart, dedicated people spending countless hours making Abby as safe, thoughtful, and effective as possible. I quickly learned that there is good AI and there is bad AI. Abby falls firmly into the good AI category — a distinction that has been earned through relentless testing, training, governance, and continuous refinement of the model.
Safety Doesn't Happen by Accident
You already know Abby was built by therapists, but what you may not know is that Abby's development is continuously monitored, evaluated, and stress-tested by therapists as well. Abby operates within a robust risk-mitigation framework and is supported by strong internal governance that oversees her ongoing development and behavior.
The truth is that many of Abby's safety and efficacy features took years to develop and refine (keeping in mind that one year in the AI world is like ten years in the human world).
In fact, the team delayed Abby's debut in the app stores until they were confident the app experience met the same high standards for safety and quality as the web platform. From the beginning, the priority has never been to be first: it has been to build one of the safest and most thoughtfully designed AI companions for emotional support.
I have been a therapist for more than 15 years, working with diverse populations across many areas of behavioral health. My integrity as a therapist is everything to me, which is why I feel so aligned with Abby. I know I'm part of a team that was willing to delay launching in the app stores for as long as it took to ensure the app met the same high standards for safety, thoughtfulness, and effectiveness as the web experience.
In the fast-moving world of AI, there is often tremendous pressure to launch quickly and improve along the way. Abby certainly continues to evolve like any AI product, but the team wasn't willing to compromise on the foundational safety and quality of the platform. Those core safeguards came first, even when delaying the app launch meant delaying revenue. As both a therapist and a member of the team, I'm proud to be part of a company that chose integrity over profit.
A New Responsibility
After reading the survey results, I'm beginning to understand that one of my roles is to help my colleagues distinguish between good AI and bad AI. Through education and transparency, I believe I can help therapists better understand what makes an AI for mental health safe, ethical, and worthy of trust.
The APA has already created an excellent guide to help clinicians thoughtfully evaluate AI-generated advice, which is an important first step. But I think our profession can go even deeper. Therapists also need a framework for understanding what AI is actually doing for clients, where it fits within the continuum of care, and what our role as clinicians becomes when clients begin incorporating AI into their emotional lives.
The Conversation We Need
I'm not very interested in debating whether this technology should exist. First, I don't think we'll truly understand the impact of AI on society until many, many years from now. Second, AI for mental health is already here. If the APA survey tells us anything, it's that people are already using it. We're not going to put the genie back in the bottle.
The conversation I'm interested in having is much more practical. I want to help therapists understand how to identify safe AI, how to think about the role AI plays in their clients' lives, and how that role can work alongside therapy. I also want to help clients understand how to use AI safely, how to recognize which tools are worthy of their trust, and the ways AI can support their mental health without replacing professional care.
Helping the Good AI Win
The stakes are high because people's mental and emotional health is on the line. That's why I believe it's so important that the good AIs like Abby win this one.
