This piece originally appeared on The Alliance for Citizen Engagement. It was written by Nicole Kwan.
Background
The American mental healthcare system struggles with deep inaccessibility. While the National Alliance on Mental Illness (NAMI) estimates that one in five adults and one in six youth experience mental illness, approximately 28 million people receive no treatment at all. Provider shortages worsen the issue, with around 123 million people living in Health Professional Shortage Areas (HPSAs). Even when services are available, access is often delayed: less than 20 percent of psychiatrists are accepting new patients, and the median wait time for an appointment is 67 days for in-person visits and 43 days for telepsychiatry. In response to these systemic barriers, many are turning to artificial intelligence (AI) to expand access to mental health care.
History
In the 1960s, Joseph Weizenbaum’s ELIZA was the first significant attempt to integrate AI into mental health, providing a basic natural language processing (NLP) system that simulated a psychotherapist. In the 1980s and ‘90s, AI gradually made improvements in mental health treatment through the introduction of expert systems. Designed to simulate human experts in decision-making, these systems were used for diagnoses and treatment planning. In 2002, researchers explored the potential of multimedia for computer-assisted cognitive therapy, expanding earlier studies’ focus on software that only displayed text. The 2010s marked the emergence of deep learning techniques, a type of AI that uses networks of algorithms to recognize patterns in large amounts of data. This allowed for more accurate predictions, diagnostics, and personalized treatments through analyses of complex data such as speech, text, and brain scans. After the COVID-19 pandemic exacerbated mental health issues globally, AI chatbots were adopted in mental health services. Today, AI is increasingly integrated into the healthcare sector, with tools such as wearable devices monitoring symptoms, identifying early signs of relapse, and providing real-time feedback to both patients and clinicians.
Regulatory Landscape
The U.S. stance on AI regulation has shifted over the years. In 2023, the Biden Administration introduced an Executive Order and a Blueprint for an AI Bill of Rights, focusing on the safe development of AI systems with principles like data privacy, protections against algorithmic discrimination, and user transparency. In 2025, the Trump Administration revoked the order, emphasizing innovation and global leadership in AI.
In the absence of comprehensive legislation, agencies such as the U.S. Food and Drug Administration (FDA) and the Federal Trade Commission (FTC) play key roles in overseeing AI in healthcare. The FDA regulates AI-enabled medical devices and digital health tools, while the FTC safeguards consumers from AI misuse. Additionally, the American Psychological Association (APA) has established ethical standards regarding the use of technology in therapy, emphasizing the need for clinician oversight in AI-assisted care. The Health Insurance Portability and Accountability Act (HIPAA) also proposed guidelines to regulate the protection of patient data when used by AI-driven mental health platforms.
Given the fragmented nature of federal oversight, state governments have taken the lead in addressing the ethical, clinical, and consumer protection challenges posed by AI in mental health care. For example, Rhode Island (HB 6285) requires licensed mental health professionals to obtain approval from the licensing board and informed consent from patients before using AI in treatment. California (S 579) proposed the establishment of a working group to evaluate the risks, benefits, and ethical implications of AI technologies in mental health services. Meanwhile, New Jersey (A 5603) prohibits companies from marketing AI as licensed mental health professionals, aiming to curb deceptive advertising and establish enforcement mechanisms through the state’s consumer fraud statutes.
Arguments in Favor
Expanding Access to Care
Mental healthcare access remains a significant challenge, particularly in rural and underserved areas with shortages of qualified professionals. High costs and long wait times further exacerbate the problem, leaving many without the support they need. To address this, AI-driven platforms have emerged as an affordable and flexible solution, helping to bridge the gap and democratize access to mental health services. For example, AI-powered chatbots provide 24/7 support, offering immediate assistance to those in distress. Crisis Text Line has integrated these chatbots to deliver coping strategies and connect users to human therapists when needed. Furthermore, AI-driven virtual therapists that use facial recognition technology can support children with autism spectrum disorder, broadening access to specialized care and helping meet the growing demand for mental health resources.
Assistance with Clinical Diagnosis
Researchers have discovered that AI has various capabilities that can improve clinical diagnoses. For one, it can help differentiate between bipolar and unipolar depression by analyzing brain imaging features. Beyond this, it can build comprehensive models of mental illness by integrating data from various sources, including electronic health records (EHRs), wearable sensors, speech patterns, and social media feeds. By analyzing longitudinal data, AI further refines diagnostic accuracy as symptoms evolve over time. In line with these advancements, Limbic, a British AI startup, developed a diagnostic tool that has evaluated over 210,000 patients, achieving a 93 percent accuracy rate in detecting the eight most prevalent mental health conditions, including depression, anxiety, and post-traumatic stress disorder.
Personalization of Care
By analyzing large datasets and utilizing machine learning, AI can assess a person’s unique characteristics, such as their “genetic predispositions, past treatment responses, behavioral patterns, and real-time physiological data,” ensuring that treatment plans are tailored to each patient. One key advantage of AI is its ability to optimize medication treatment based on genetic factors. Through reviewing a patient’s genetic profile, AI can predict how they will respond to specific antidepressants, resulting in more effective treatments and fewer adverse side effects. Additionally, by continuously monitoring a patient’s behavioral patterns, AI can help therapists refine their approach and customize interventions based on each patient’s specific challenges and strengths. In cognitive behavioral therapy (CBT), for example, AI could pinpoint perfectionism as a key issue for the patient, thereby improving the effectiveness of therapy.
Arguments Against
Data Privacy and Security Issues
The use of AI in mental health care raises concerns about the commercialization of sensitive data and the lack of regulatory oversight. A Politico article reported that Loris AI, the for-profit spin-off of Crisis Text Line, uses anonymized data from the nonprofit to improve its customer service software. However, experts caution that even anonymized data could potentially be re-identified, violating patients’ privacy. Additionally, there is a broader issue in federal regulations; mental health nonprofits are not subject to the same level of oversight as for-profit companies, despite handling equally sensitive data. BetterHelp illustrates the danger of such a gap in data privacy protection. In 2023, the FTC took action against the company for sharing users’ personal information with major tech companies, who then used that data for targeted advertising, despite BetterHelp’s prior assurance that the information would remain confidential.
Bias and Lack of Cultural Sensitivity
Since psychological research has historically focused on predominantly White and high-income populations, the datasets NLP models are trained on often reflect biases. Researchers found that this lack of diversity limits the socio-cultural sensitivity and inclusivity of mental health treatments and thus leads to inaccurate diagnoses and treatment recommendations. Language models (LLMs) are also prone to misinterpreting cultural behaviors and expressions of distress, often mistaking them for symptoms of mental disorders. For example, a 2025 study found that LLMs provided significantly different psychiatric treatment recommendations when a patient’s race was implied or explicitly stated.
Liability and Social Risks
A major issue for AI-based mental health services is the potential for addiction. Many chatbots employ techniques similar to social media platforms that trigger dopamine, leading to dependency and social isolation. Research has found that students who use chatbots for emotional support are more socially disconnected, particularly when relying on them during moments of loneliness. This suggests that these interactions may not only fail to address one’s mental state, but could potentially worsen it. Furthermore, AI chatbots have been discovered to misrepresent themselves as licensed mental health professionals. These deceptive practices are particularly dangerous for vulnerable groups like children and adolescents, as highlighted by a recent TIME investigation that revealed alarming interactions with general-purpose chatbots like Replika and Nomi. One bot encouraged a suicidal teen to join it in the afterlife while another attempted to quell a 15-year-old boy’s violent urges by suggesting an “intimate date” with the chatbot itself. These examples underscore the blurry line between accountability and technological advancements.
Future Outlook
Nearly 80 percent of U.S. adults stated that they would not use an AI chatbot for mental health support, reflecting concerns about the ethical implications of AI. One key challenge in AI-driven mental health care is obtaining informed consent, since certain conditions may temporarily impair decision-making and complicate the ethical use of data. As AI becomes more embedded in mental health care, clinicians may need to improve their digital literacy skills and receive training to effectively oversee or collaborate with these tools. However, this shift may also spark concerns about job displacement in the mental health profession. Looking ahead, prioritizing thoughtful oversight, ethics, and human-centered policies will be key to ensuring that AI develops in ways that support, rather than replace, quality mental health care.
