Artificial intelligence has moved from being something people associate with technology and productivity to something increasingly used for emotional support. People now turn to AI chatbots to talk through difficult feelings, manage stress, understand their thoughts, practice coping strategies, and sometimes even explore symptoms of anxiety or depression.

That raises an important question: Does AI therapy actually work?

The answer from clinical research is promising, but it is not as simple as saying yes or no.

Studies suggest that some AI-based mental health interventions can reduce symptoms of depression, anxiety, stress, and psychological distress. However, researchers also emphasize that the evidence varies significantly between different technologies. A purpose-built mental health chatbot that has been tested in clinical research is not necessarily comparable to a general-purpose generative AI chatbot.

For anyone considering AI therapy, understanding that distinction is essential.

What Does “AI Therapy” Actually Mean?

The term AI therapy is used broadly, which can make the research difficult to interpret.

Some tools are specifically designed around psychological techniques such as cognitive behavioral therapy (CBT). Others use conversational AI to provide general emotional support, journaling prompts, psychoeducation, or coping suggestions.

There is also a major difference between traditional rule-based mental health chatbots and newer generative AI systems that create responses dynamically.

This matters because the research supporting one type of technology should not automatically be applied to another.

The American Psychological Association (APA) has specifically cautioned that general-purpose generative AI chatbots have not been sufficiently studied to establish that they can safely provide psychological treatment. At the same time, the APA notes that some purpose-built mental health technologies have shown promising results in research. 

What Have Clinical Studies Found?

One of the most important developments in AI therapy research has been the increasing number of randomized controlled trials.

These studies are particularly valuable because they allow researchers to compare an intervention with another condition, such as information-only support or a control group.

Early Research With Woebot

An early randomized controlled trial investigated Woebot, a conversational agent designed around CBT principles.

The study included 70 young adults between 18 and 28 years old who reported symptoms of anxiety or depression. Participants were randomly assigned to either use the conversational agent or receive an information-only resource.

After two weeks, researchers found a significant reduction in depression symptoms among participants using the conversational intervention compared with the information-only group.

The study was relatively small and short, so it cannot establish that AI therapy works for everyone. However, it provided early evidence that a structured conversational intervention could produce measurable changes in some mental health symptoms.

What Do Larger Reviews Tell Us?

Individual studies are useful, but systematic reviews and meta-analyses provide a broader picture because they combine evidence from multiple studies.

A 2024 systematic review and meta-analysis examined 18 randomized controlled trials involving 3,477 participants and evaluated AI-based chatbots for depression and anxiety.

The researchers found small but statistically significant improvements:

• Depression: effect size of approximately −0.26
• Anxiety: effect size of approximately −0.19

The researchers also found that benefits were more apparent after several weeks of intervention, while some effects were not maintained at longer follow-up periods.

This is an important finding. It suggests that AI-based interventions may help with certain symptoms in the short term, but researchers still need stronger evidence about whether these improvements last.

Research Among Young People Is Also Growing

Research has increasingly examined AI mental health tools among adolescents and young adults.

A recent systematic review and meta-analysis covering 31 randomized controlled trials and 29,637 adolescents and young adults found that AI chatbot interventions were associated with improvements in several measures of mental distress.

The reported effects included improvements in depression, anxiety, and stress.

However, large participant numbers do not automatically eliminate all limitations. Different studies may use different chatbot technologies, therapeutic approaches, populations, and outcome measurements.

For this reason, researchers continue to emphasize the need for better-designed trials and longer follow-up periods.

Why Might AI Therapy Help?

One reason AI-based support is attractive is simple: accessibility.

Traditional mental health care can involve waiting lists, geographic limitations, financial barriers, transportation problems, or difficulty finding a professional who is accepting new clients.

AI tools can be available at almost any time.

That accessibility may be particularly useful for people who are hesitant to make their first appointment with a therapist.

AI can also make certain mental health activities easier to repeat. For example, someone might use a conversational tool to:

• Reflect on stressful situations
• Identify recurring thought patterns
• Practice coping techniques
• Keep track of mood changes
• Learn about common mental health concepts
• Prepare notes or questions for a therapy session
• Build healthier daily habits

APA guidance published in 2026 notes that AI may be useful for organizing thoughts, generating questions, and prompting reflection, particularly when it is used alongside professional mental health care. 

The Human Side of AI Therapy

There is another reason people may find conversational AI appealing: it can feel easier to talk to a machine.

Someone may be embarrassed to tell another person about a particular thought or worry. A chatbot does not visibly react, interrupt, or appear to judge.

That can lower the psychological barrier to starting a conversation.

But there is an important distinction between simulating empathy and providing professional therapeutic care.

An AI system can generate language that sounds understanding and supportive, but it does not have the clinical judgment, personal responsibility, or human relationship of a licensed therapist.

The therapeutic relationship itself is an important part of psychotherapy. A trained clinician can understand a person’s history, observe changes in behavior and communication, evaluate risk, and adjust treatment over time.

AI does not currently provide that same level of clinical understanding.

What Psychologists Are Seeing in Real-World Practice

Research isn’t limited to clinical trials. Mental health professionals are increasingly encountering AI use among their patients.

In its 2026 Chatbots and Mental Health Survey, the APA surveyed more than 1,200 licensed psychologists in the United States.

The survey found that 77% of psychologists had spoken with patients who had used AI for support, engagement, or other reasons. About half reported seeing positive communication with chatbots, while 25% reported seeing unhealthy communication. 

At the same time, psychologists expressed considerable concern about safety.

For example:

• 94% reported concerns about some patient use of chatbots.
• 97% felt chatbots could inadvertently reinforce negative behaviors or delusional beliefs.
• 89% were concerned that chatbots could inadvertently encourage self-harm.
• 36% reported noticing some level of chatbot dependency among patients.

These numbers describe psychologists’ observations and concerns rather than proving that AI directly causes these outcomes. Nevertheless, they show why professional organizations are urging caution.

AI Therapy Is Not the Same as Traditional Therapy

This is perhaps the most important point when interpreting the research.

Clinical evidence showing that a particular AI intervention can reduce symptoms does not mean that AI has replaced psychotherapy.

A therapist can do things that current AI systems cannot reliably do.

For example, a therapist can:

• Conduct a comprehensive clinical assessment
• Evaluate complex personal and family history
• Recognize subtle behavioral and emotional changes
• Assess suicide or self-harm risk
• Adjust treatment based on progress
• Challenge unhealthy patterns when appropriate
• Coordinate care with other healthcare professionals
• Take professional and ethical responsibility for treatment

AI can provide useful support around some of these activities, but it should not be assumed to perform them at the same level.

The APA currently recommends that AI chatbots and wellness applications be viewed as supplementary tools rather than replacements for qualified mental health professionals. 

What About Generative AI?

This is where people need to be especially careful.

A growing number of people use general-purpose systems such as generative AI chatbots for emotional support. However, the clinical evidence for these systems is much less established than the evidence for some purpose-built mental health interventions.

The APA’s health advisory states that general-purpose generative AI systems have little to no established evidence base for mental health treatment and may lack adequate safety protocols, clinical oversight, and scientific validation.

This does not mean every interaction with a general AI system is harmful.

It means that users should not confuse a fluent, empathetic-sounding conversation with clinically validated treatment.

AI can sound confident even when it is wrong.

Where AI Therapy May Be Most Useful

Based on the research available so far, AI may be most appropriate for specific, limited purposes rather than attempting to replace comprehensive psychotherapy.

For example, it may support:

Self-Reflection

AI can provide prompts that encourage someone to examine thoughts, emotions, behaviors, and recurring stressors.

Coping Practice

Some purpose-built tools can guide users through structured techniques derived from psychological approaches such as CBT.

Between-Session Support

People already seeing a therapist may use digital tools to track their mood, organize thoughts, or prepare topics for their next appointment.

Mental Health Education

AI can help explain basic concepts and encourage people to learn more about their mental health.

Taking the First Step

For someone who has never considered therapy, a conversational tool may provide an accessible starting point before seeking professional help.

These uses are more consistent with current professional guidance than treating an AI chatbot as a replacement therapist.

What the Research Still Doesn’t Tell Us

Despite encouraging findings, there are still major unanswered questions.

Researchers need more evidence about:

• Long-term effectiveness
• Safety over extended periods
• Different age groups
• Cultural differences
• Severe mental health conditions
• Crisis situations
• Emotional dependence on AI
• Privacy and data protection
• Generative AI compared with structured chatbots
• How AI should be integrated with professional treatment

The APA has emphasized that the rapid development of AI has moved faster than the research needed to evaluate its effects and recommends more rigorous, transparent scientific investigation.

How Should People Use AI Therapy?

A sensible approach is to think of AI as a mental health support tool rather than a replacement for human care.

It may be useful for reflecting on your day, learning coping strategies, organizing questions before a therapy appointment, or tracking emotional patterns.

But avoid relying on AI to diagnose a mental health disorder, interpret psychological tests, change medication, or manage an emergency.

If symptoms are persistent, worsening, or interfering with your relationships, work, school, sleep, or everyday life, speaking with a qualified mental health professional is more appropriate.

And if someone is experiencing suicidal thoughts, self-harm urges, psychosis, or another mental health emergency, they should seek immediate professional or emergency assistance rather than relying on an AI chatbot.

So, What Does Clinical Research Actually Say?

The research is promising but still developing.

Clinical trials and meta-analyses suggest that certain AI-based conversational interventions can produce measurable improvements in depression, anxiety, stress, and psychological distress. However, the effects are generally modest, vary between interventions, and are not always sustained over longer follow-up periods.

Just as importantly, the evidence for purpose-built mental health chatbots should not be treated as evidence that every generative AI chatbot is an effective therapist.

The most realistic role for AI right now is as an accessible supplement to mental health support. It can help people reflect, practice certain coping techniques, organize their thoughts, and potentially take the first step toward professional care.

The technology is developing quickly. The science needs to keep pace.

References

1. American Psychological Association. Health Advisory: Use of Generative AI Chatbots and Wellness Applications for Mental Health. APA Health Advisory
2. American Psychological Association. Patients Are Bringing AI to Therapy: 2026 Chatbots and Mental Health Survey. APA 2026 Survey
3. American Psychological Association. Guide to Navigating AI-Generated Advice Thoughtfully and Safely. APA AI Guidance
4. Fitzpatrick, K. K., Darcy, A., & Vierhile, M. Delivering Cognitive Behavior Therapy to Young Adults With Symptoms of Depression and Anxiety Using a Fully Automated Conversational Agent (Woebot): A Randomized Controlled Trial. JMIR Mental Health.
5. Zhong, W., Luo, J., & Zhang, H. The Therapeutic Effectiveness of Artificial Intelligence-Based Chatbots in Alleviation of Depressive and Anxiety Symptoms in Short-Course Treatments: A Systematic Review and Meta-Analysis. Journal of Affective Disorders, 2024.
6. Feng, X., et al. The Effectiveness of AI Chatbots in Alleviating Mental Distress and Promoting Health Behaviors Among Adolescents and Young Adults: Systematic Review and Meta-Analysis. Journal of Medical Internet Research, 2025.