A Harvard Business Review survey has announced, for a second year in a row, that “therapy and companionship” is the number one use case of generative artificial intelligence (AI), based on nearly 50,000 posts drawn from LinkedIn, TikTok, YouTube, Reddit, and other platforms.
The term “therapy” in this case refers to individuals using generative AI for emotional support, which is not the equivalent of psychotherapy delivered by a licensed mental health professional, but may reflect the broader notion that users experience engagement with large language models “as if” it is therapy. This offers the sense of being in therapy, but it is de facto or quasi-therapy. The emotional use of AI is sometimes referred to as “affective use,” but this does not fully capture the specific relationship of attachment and authoritative positioning of AI models playing therapist-like roles.
We do not yet have a distinct and separate term to refer to such conversations with AI that feel therapy-like to the user but are not delivered within a therapeutic frame.
I propose a new term, “para-therapy,” to capture this emotional and relational engagement with generative AI, where users relate to AI as therapists, explicitly or implicitly, consciously or unconsciously. Para-therapy, while not the same as psychotherapy, can satisfy and fulfill user expectations of a therapy-like experience and even allow people to feel better. Such interactions with general-purpose AI chatbots may even lower depressive symptoms, but para-therapy lacks the clinical infrastructure, stable therapeutic frame, informed consent, and ethical boundaries that ensure the safety and effectiveness of psychotherapy. The question is not whether conversational AI can sometimes help improve mood or anxiety—it often can—but whether such help is the same as targeted therapy, and whether we are comfortable erasing that distinction.
Para-therapy Is Becoming More Common—Often in Secret
In a new study by McBain and colleagues, nearly one in five adolescents and young adults in the United States use AI chatbots for mental health advice. Ninety-two percent of them said advice was helpful. But nearly two-thirds said that they had not told anyone.
With an estimated 25 percent of US adults turning to general-purpose AI chatbots for emotional support, we are witnessing a deeper paradigm shift: In a process of reverse influence, these large language models occupying therapy-like roles are shaping what counts as “therapy” and expectations of human therapists, especially for those who may not have previously been in psychotherapy with a mental health professional. Para-therapy can also impact and potentially disrupt existing therapeutic relationships, creating a triangulation or splitting effect if one already has a human therapist.
Some argue that large language models are filling the gap with something, even if it is not “therapy”—and that their scalability and affordability are their advantages. The question of “Is something better than nothing?” continues to be a recurring refrain. Of note, para-therapy, while useful for many, can carry downstream risks, especially for those with more acute or severe symptoms that should be addressed by mental health professionals.
Is “AI Para-therapy” Actually Therapy?
General-purpose chatbots and AI companions are not intended to provide therapy but can provide interactions that feel helpful. General emotional support, empathic responses, personalized advice and resources, and psychoeducation are being equated with therapy, but are not the equivalent of psychotherapy. The fact that such AI chatbots can simultaneously occupy many other relational roles for the user—tech support, intimate partner, editor—already signals that a fixed therapeutic frame does not exist. Less clear are specialized AI chatbots that are narrowly designed and marketed for mental wellness and mental health, including digital therapeutics, and whether we feel comfortable calling those therapy versus para-therapy. There are hundreds of different evidence-based therapies—some that may be more effectively delivered by a conversational AI—and not all modalities are the right clinical match for a situation.
The Regulatory Gap
The widespread use of general-purpose AI models for emotional support currently sits in a regulatory gap, a fragmented and largely unregulated space. Even AI chatbots that make claims about improving well-being have limited regulatory oversight, which has raised the analogy to the nutritional supplement or yoga wellness industry, in a study by Cooper and colleagues.
This puts general-purpose chatbots and AI companions outside the oversight and requirements of traditional therapy, which involve professional licensure, state licensing boards, mandated reporting requirements, duty-to-warn obligations, federal (HIPAA) and state privacy protections, and ethical guidelines issued by professional organizations.
What Will Become of Therapy?
This shift toward using general-purpose AI chatbots for mental health has led some in the field to warn that AI will soon dominate psychotherapy. Like many other professions, there are fears of labor replacement within the therapist community. In the most recent Stanford Index Report on AI, about one in five AI experts and therapists believe that mental health therapists will experience labor replacement in the next 20 years.
What is uncertain is whether users are consciously treating AI as therapists, unconsciously relating to AI in a fluid role similar to therapists, or are actively aware that it is not therapy but approximates it well enough. This situation is forcing therapists to confront how the work is distinct from merely offering language, cognitive insights and interpretations, and a private space for emotional support.
Para-therapy may fill a gap; psychotherapy is distinct, for now. Therapists offer the human capacity to be present, sit with uncertainty, sense and deliver well-timed challenges, track and align with overarching therapeutic goals, and operate and perceive beneath and beyond language to facilitate attunement, co-regulation, and meaningful silence. What is also rarely acknowledged is that, in many cases, the therapeutic goal is to come to terms with our own and others’ limitations as humans, including the limitations of therapists. Large language models may be able to explain and even simulate this, but the experience of navigating boundaries, the friction of rupture and repair, and the emotional processing of such disappointment with another human offers a unique path to healing.
Marlynn Wei, MD, PLLC, Copyright 2026. All Rights Reserved.
To find a therapist, visit the Psychology Today Therapy Directory.