The APA recommends: “Do not rely on GenAI chatbots and wellness apps to deliver psychotherapy or psychological treatment.” It also says these tools “should not be used as a replacement for a qualified mental health care provider, but may be appropriate as a supportive adjunct, not substitute, to an ongoing therapeutic relationship.”

WHO materials on mental disorders emphasize diagnosis and treatment within health systems and trained care, not self-treatment by automated chatbots. This provides global context that mental health conditions are clinical disorders requiring appropriate professional care.

APA is urging federal regulators to implement safeguards against AI chatbots posing as therapists, warning that unregulated mental health chatbots can endanger the public. The page says chatbots can help in some gaps when therapists are unavailable, but it also warns they may discourage people from seeking a trained human therapist and, in extreme cases, encourage harm.

Reuters reporting in 2025–2026 has covered concerns that AI chatbots can give unsafe or inaccurate mental-health advice and that experts do not view them as substitutes for therapy. The reporting also notes that some tools can quickly direct distressed users to resources.

The APA says AI can be useful for organizing thoughts, exploring evidence-based coping strategies, preparing for appointments, and generating ideas for discussion with a clinician. It emphasizes that AI should not replace care from a licensed mental health professional and advises against relying on AI to diagnose mental health conditions or provide support during a mental health crisis.

The meta-analysis of 14 RCTs showed a statistically significant effect (effect size [ES] = 0.30, p = .047, N = 6314, 95% CI 0.004, 0.59), which means that GenAI chatbots are, on average, effective in reducing negative mental health issues, such as depression, anxiety, among others. This corresponds to a small-to-moderate positive effect of chatbot-based mental health interventions compared to control groups. It seems like AI chatbots can assist in-person counseling sessions but might not replace human therapy.

Across the 14 included studies, CBT-based chatbots consistently demonstrated short-term reductions in depressive symptoms, whereas findings for anxiety symptoms were mixed: some trials reported improvements, whereas others found no significant effects or did not report anxiety outcomes. Preliminary findings from randomized controlled trials (RCTs) suggest that CBT chatbots can yield significant reductions in depression and anxiety over short-term interventions (typically 2‑8 weeks), particularly when compared to passive or information-only controls. While results against active comparators (eg, human-guided therapy or coaching apps) were mixed, these tools consistently demonstrated moderate effect sizes for depressive symptoms, underscoring their potential as first-line or supplementary digital interventions, rather than clear replacements for human therapists.

NIH resources distinguish evidence-based mental health treatment from general wellness tools and emphasize evaluation of safety, efficacy, and clinical oversight. This supports the distinction between supportive digital tools and actual therapy delivered by trained clinicians.

This systematic review and meta-analysis included randomized controlled trials comparing chatbots with any control condition for depressive and/or anxiety outcomes. Of the 39 eligible studies, 38 (n = 7,401) were analyzed for depression and 34 (n = 7,621) for anxiety. Chatbots produced statistically significant reductions in depressive (Hedges' g = 0.31, 95% CI [0.17, 0.46]) and anxiety symptoms (g = 0.28, 95% CI [0.05, 0.51]) compared with controls. The authors conclude that contemporary chatbots appear to alleviate depressive and anxiety symptoms, especially in individuals with greater depressive severity, but the trials largely compare against control conditions and do not establish that chatbots can fully replace human therapists.

Fifty-four articles were chosen for further review, with 10 articles included in the final analysis. Overall, most studies showed positive trends in improving anxiety, stress, and depression. Overall, using an artificial intelligence chatbot for mental health has some promising effects. However, many studies were done using rudimentary versions of artificial intelligence chatbots. In addition, lack of guardrails and privacy issues were identified. More research is needed to determine the effectiveness of artificial intelligence chatbots and to describe undesirable effects.

The report states that patients are turning to artificial intelligence and chatbots to supplement their relationships with licensed mental health professionals. It frames AI as a supplement to therapy, not a replacement for licensed care.

GenAI chatbots were not created to deliver mental health care, and wellness apps were not designed to treat psychological disorders, but both technologies are increasingly being used by consumers for these purposes. These tools are not a replacement for a psychologist or other trained mental health professional. They should not be used for diagnosis or treatment of mental disorders, and they are not appropriate for crisis intervention.

In this randomized clinical trial, participants were assigned to use a conversational AI agent or receive usual care. The generative AI system produced symptom reductions in depression and anxiety compared with a waiting list control, with significant improvements in psychiatric symptoms and functioning over 8 weeks. However, the study compares the AI agent to waiting list or minimal intervention rather than to full human-delivered psychotherapy, and the authors emphasize that clinician oversight and integration into standard care remain important.

“There is absolutely no consensus in the field that AI chatbots can serve in any way as a replacement for therapy.” – C. Vaile Wright. Understanding what AI chatbots can and cannot do is critical for understanding their potential, Girgis noted, and what they do is mirror the data or emotions shared with them. AI chatbots can provide resources, but cannot replace therapy or crisis intervention services. Chatbots are not qualified to address crises themselves but often attempt to do so, potentially harming users, and have been documented in engaging in extremely harmful behaviors like sharing lethal means of suicide and coaching users to hide their mental health symptoms.

This peer-reviewed study found clinicians saw benefits such as homework support, multilingual access, affordability, and 24/7 availability. It also found major risks including lack of regulation, privacy concerns, overreliance, incorrect treatment recommendations, and inability to detect subtle cues such as tone and eye contact.

Dartmouth researchers conducted the first-ever clinical trial of a generative AI-powered therapy chatbot and found that the software resulted in significant improvements in participants’ symptoms, according to results published March 27 in NEJM AI. People diagnosed with depression experienced a 51% average reduction in symptoms, and participants with generalized anxiety reported an average reduction of 31%, with many shifting from moderate to mild anxiety. The senior author stated: “The improvements in symptoms we observed were comparable to what is reported for traditional outpatient therapy… Our results are comparable to what we would see for people with access to gold-standard cognitive therapy with outpatient providers,” while also noting that AI-powered therapy is still in critical need of clinician oversight.

The study found that mental health apps produced small but statistically significant improvements in symptoms of depression (g = 0.28) and generalized anxiety (g = 0.26) across tens of thousands of participants. For anxiety, this study found that the treatment with therapy chatbots resulted in significant reductions of anxiety symptoms (p < 0.001) with an effect size of g = −.19 overall. However, it is important to note that at 3‑month follow up this effect size was diminished and nonsignificant, indicating a potential limitation in such interventions. Though therapy chatbots appear to be equal in effectiveness at this time, there are several interesting considerations that indicate therapy chatbots may eventually prove to be more effective than smartphone mental health applications.

This randomized controlled trial evaluated an AI therapy chatbot as a self-help intervention for university students with depressive symptoms. The therapy chatbot reduced depression in the 16‑week intervention period and reduced anxiety in the first 4 weeks. A significantly better therapeutic alliance was reported in the chatbot group compared with the control, indicating that users felt able to form a working relationship with the AI. Nonetheless, the study positions the chatbot as a self-help tool and does not claim that it can replace a human therapist for treating mental health conditions.

A new Stanford study reveals that AI therapy chatbots may not only lack effectiveness compared to human therapists but could also contribute to harmful stigma and dangerous responses. Low-cost and accessible AI therapy chatbots powered by large language models have been touted as one way to meet the need. But new research from Stanford University shows that these tools can introduce biases and failures that could result in dangerous consequences. While using AI to replace human therapists may not be a good idea anytime soon, Moore and Haber do outline in their work the ways that AI may assist human therapists in the future.

The design of generative AI models make them poor substitutes for mental healthcare professionals, caution TC experts. There have been reports of people with no history of mental illness experiencing delusions brought on by the chatbot, and multiple teenagers have died by suicide while engaged in relationships with AI companions. Research has shown that when AI chatbots, including ChatGPT, were given prompts simulating people experiencing suicidal thoughts, delusions, hallucinations or mania, the chatbots would often validate delusions and encourage dangerous behavior. “The conclusions of that study were clear, the potential for serious harm meant AI is simply not ready to replace a trained therapist, at least not yet,” says Nitzburg.

Brown researchers reported that AI chatbots, even when prompted to use evidence-based psychotherapy techniques, systematically violate ethical standards of practice. The study identified harms including poor crisis handling, misleading responses that reinforce negative beliefs, and a false sense of empathy.

The APA Monitor reports that randomized trials have found some mental health chatbots can reduce symptoms of depression and anxiety, especially over short periods and compared with waitlist or information-only controls. Psychologists interviewed note that these tools may be useful for people who do not have access to therapy and can provide coping skills or psychoeducation. However, the article stresses that chatbots "cannot replace a trained therapist," may miss nuances such as suicidality or complex trauma, and raise concerns about accountability, data privacy, and the therapeutic relationship.

Stanford HAI reported that a new Stanford study found AI therapy chatbots may lack effectiveness compared with human therapists and could contribute to harmful stigma and dangerous responses. The article says some people see real benefits, but it highlights clear safety concerns.

As global demand for mental health services surges, AI-powered chatbots and apps are already proving to be invaluable assets in the provision of evidence-based psychotherapy such as cognitive behavioral therapy, breaking down traditional barriers to treatment such as cost, access, stigma, and therapist availability and fit. Evidence of AI's efficacy in mental health treatments is growing, with health systems like the U.K.'s National Health System (NHS) employing AI apps to handle an overwhelming number of referrals for mental health services. AI therapy can also relieve the strain on human therapists, who themselves handle the mental and emotional load of serving many distressed people each day.

The project description states that digital interventions powered with generative artificial intelligence have shown promise in building personalized, on-demand support at scale. Promising results from a randomized controlled trial (N = 210) using the generative-AI-driven mobile app Therabot demonstrated high anxiety and depression symptom reduction compared to control, with high engagement averaging 600 messages and 6 hours of use per participant. The authors emphasize the "great potential benefit" of such interventions but note it is currently unknown whether Therabot can be used to treat co-occurring substance use disorders and describe plans for further trials, implying that the technology is still under investigation and not yet a clear substitute for standard therapist-led care.

These chatbots can actively listen, provide emotional support, and even deliver cognitive‑behavioral interventions. Individuals gain access to a broader range of mental health resources and support services, which can be particularly beneficial for addressing conditions such as depression, anxiety, and stress. However, despite these promising capabilities, concerns remain regarding the clinical validation, safety, and ethical implications of deploying AI chatbots as stand‑alone mental health treatments.

Nearly two in five psychologists (39%) reported patients using AI tools to self-diagnose, and others are relying on chatbots for affirmations and support. The article describes this as growing use of AI alongside professional care rather than evidence that chatbots can replace therapists.

APA says its science-based guidance is designed to help people use these tools more safely and critically. It adds that feeling understood by a chatbot is not the same as being clinically understood by a trained professional who can assess risk, history, behavior, and context.

In mainstream clinical guidance, AI chatbots are generally discussed as adjunct tools for psychoeducation, journaling, or symptom tracking, not as replacements for licensed therapists. Risk concerns commonly cited include hallucinated advice, privacy issues, inability to assess suicide risk reliably, and lack of accountability or therapeutic alliance.

This post says the APA, in partnership with its Mental Health Technology Advisory Committee, developed ethical guidance for AI in the professional practice of health service psychology. It presents the resource as designed to ensure emerging technologies in behavioral health are safe, effective, responsible, and ethical.