Since its introduction in the 1960s, artificial intelligence (AI) has permeated modern life in ways largely unimagined. With nearly 1.8 billion people around the world utilizing AI for daily conveniences to complex professional projects, it is a cornerstone of current societal infrastructure. However, the promise of AI for increased administrative efficiency has brought forth unintended consequences for the rising global mental health crisis.
To exemplify, in the U.S., about one in eight adolescents and young adults is turning to AI chatbots for mental health advice. At its core, human connection in therapy is anchored in vulnerability and empathy that allows one to feel as though they are truly seen and held by another being. As such, the transformation of AI into a surrogate for human connection poses significant ethical questions. Key concerns regarding the absence of emotional attunement and the inability of AI tools to appropriately respond to crisis circumstances challenge the perceived benefit of AI incorporation in mental health.
Mental health conditions are significant contributors to global health disorders, including anxiety, depressive, mood, psychotic, and eating disorders that require treatments ranging from prescription drugs to psychotherapies. For Americans seeking mental health support, the high costs of therapy sessions and strict insurance limitations impose high financial barriers, framing such support systems as a privilege. Without insurance, the average cost for therapy sessions in the U.S. is estimated to be $143.26. As sessions are often provided weekly, the average total cost of therapy per month, $573.04, is over 10% of the average monthly income in the U.S., which stands at $5,114. As a result, uninsured, low-income individuals face systemic barriers to mental health care.
Amidst financial barriers and a severe shortage of mental health practitioners, millions are now turning to AI models for primary emotional support. For instance, users may utilize these models to de-escalate spiraling thoughts with prompts such as: I can’t sleep, my heart is racing, and I feel like I’m losing control of everything in my life right now. In response, the chatbots may offer words of reassurance, guiding the user through a meditation exercise designed to ground them in reality. Compared to the three-digit prices of therapy, AI therapy costs range from $0 to $50 per month for unlimited access. Additionally, as AI platforms are available 24/7 without the need for an appointment or a waiting list, they alleviate the financial and temporal constraints presented by therapy.
These benefits were especially recognizable during the COVID-19 pandemic when in-person therapy was severely restricted. In response to the deteriorating mental health of healthcare workers, in July 2020, the national tertiary healthcare division in Singapore piloted the use of Wysa, a mental health mobile app featuring an AI chatbot with personalized support. With 80.1% of users completing at least two sessions, the mean engagement averaged about 10.9 sessions over 3.8 weeks. Wysa was particularly successful in alleviating sleep disturbances, irritability, and low energy associated with moderate anxiety and depression disorders. Furthermore, in a separate study on Wysa, researchers found an improvement in users’ self-reported state of mental health that was comparable to the improvements found from therapy.
Additionally, researchers at Dartmouth developed an AI therapy chatbot called “Therabot” that was trained to utilize evidence-based therapeutic techniques, exemplifying the benefits of AI integration in therapy. The clinical trial, consisting of 106 people from across the U.S. who are diagnosed with major depressive disorder, generalized anxiety disorder, or an eating disorder, interacted with Therabot. After four weeks of treatment, the results indicated a significant reduction in major depressive disorder, generalized anxiety, and eating disorder symptoms.
This data demonstrates a momentous potential that AI tools have in bridging the gap that often constrains therapy to physical offices. Moreover, the study participants remarked that their trust and communication with Therabot was comparable to working with a human therapist.
Despite its benefits, substituting AI for licensed therapists poses significant risks, as the core of AI’s operation carries an incapacity for ethical judgment, emotional depth, and clinical expertise.
One of the strongest predictors of positive outcomes across therapy sessions is the trusted relationship between therapist and client. While AI chatbots appear to provide personal insights, their responses are generated by their ability to predict the appropriate next words in the conversation’s pattern. Furthermore, AI tools inherently lack empathy, stopping short of fostering deep healing. Therapists undergo years of education and training to “recognize subtle warning signs, ask the right questions, and implement evidence-based interventions.”
AI integration in therapy overlooks the important role that emotional attunement plays in human interactions. In therapy, therapists take into account the words the client speaks and the subtle shifts in the body language. While AI’s responses may sound empathetic, they fail to register such non-verbal cues. Without such insight, AI blurs the line between convenience and competency. Although AI models, such as Wysa and Therabot, are capable of reducing symptoms of mental health disorders, their inability to produce deep healing that comes with such reductions was a clear portrayal of their limitations. To exemplify, a study conducted by Manipal University, Jaipur, India, demonstrated how users’ appreciation for AI chatbots’ convenience failed to override their dissatisfaction with their lack of meaningful emotional connection.
Furthermore, lacking contextual awareness, AI has an elevated risk of providing harmful responses. OpenAI, the AI research company behind ChatGPT, acknowledged that AI chatbots can raise safety concerns in areas of “mental health, emotional over-reliance, or risky behavior.” AI’s tendency to be sycophantic, or overly affirmative, was observed in multiple studies, particularly in its propensity to validate suicidal ideation, delusions, or harmful thinking. Research conducted at Stanford University found that AI often fails to recognize suicidal intent.
Unfortunately, these detrimental effects have already manifested into real-life tragedies. In 2025, 16-year-old Adam Raine lost his life to suicide. Investigations into his conversation with ChatGPT revealed how Adam had confided about his suicidal ideations, after which the chatbot discouraged him from seeking help and even offered to write his suicide note. While others may point to the established safeguards that prevent individuals from utilizing AI platforms in ways that encourage self-harm, various psychological and framing techniques, such as educational disclaimers, make it easy for users to bypass such safeguards. This tragedy speaks to the lack of accountability and the detrimental consequences that arise from AI incorporation as a substitute for therapy. The validating nature of ChatGPT fails to challenge the user’s destructive narratives, leading to a reinforcement of harmful ideations.
The benefits and dangers of AI integration in therapy pose ethical questions concerning its application. While a complete replacement of human therapists may bring catastrophic results, AI could be used to assist human therapists. For example, AI tools could help patients in less critical scenarios as valuable assets, allowing clients to engage in their care 24/7 without incurring the high costs of therapy.
The benefits of accessibility and low costs could transform mental healthcare, while stories of caution demand critical thinking about the precise role AI tools can play. Its inability to provide genuine connections that stand at the core of mental health therapy, and its incapability to correctly assess the severity of situations, place clear limitations on its incorporation in treating critical mental health conditions. As a society, we must keep a careful leash on the usage of AI in the mental health space, but in utilizing its advantages, we can ensure that everyone receives the personalized and holistic mental health treatments they need.
