AI Chatbots As Therapists

AI in Mental Health Care and Therapy

Artificial intelligence (AI) companions and therapy chatbots, commonly referred to as therabots, are powered by large language models (LLMs), such as ChatGPT or Gemini, and marketed as mental health companions. AI therapy chatbots respond in similar ways as a real therapist would when users come to them with problems or stressors associated with depression, anxiety, or other mental health disorders.

While these therabots do have their benefits, experts warn that these unregulated sources can lead to harmful and dangerous outcomes in those with severe mental health disorders.

How Does AI Therapy Work?

When a person interacts with and describes their situation to a therabot, the therabot asks questions, reflects on what the user says, and helps them work through issues in real time. They can suggest appropriate coping mechanisms, such as breathing exercises, journaling, and words of affirmation, and even track the user’s mood.

However, many users do not realize that most LLMs are centered on being excessively eager to please. These therabots are programmed to learn more about the user’s preferences each time they engage, then respond accordingly. Companies that operate these chatbots generate revenue from continued use and repeat visits, so chatbots are designed to form a strong bond with users, to the point where users may even become addicted. As a result, an AI therapist will tend to be more agreeable, eager to please, and provide needless validation compared to a human therapist.

Additionally, therapy chatbots cannot make a mental health diagnosis or prescribe medications and are not held to the same ethical guidelines. For example, HIPAA does not apply to AI therapy chatbots, and they are not subject to FDA regulation. Therefore, it is essential to remember that these chatbots do not possess the same training or ethical standards as human therapists.

Research on AI Therapy

Research on the effectiveness of AI therapy is limited because this area is still relatively new. However, early findings suggest that AI chatbots can be effective in addressing symptoms of mental health disorders and that users generally experience positive outcomes.

ChatGPT has approximately 700 million weekly users, with over 10 million paying $20 per month for the paid subscription version. A study from March 2025 found that 48.7% of participants self-reported using major AI chatbots to assist with their mental health challenges. The study also found:

• 73% use chatbots for anxiety management, 60% for depression support, 58% for emotional insight, 56% for mood improvement, and 35% to feel less lonely.
• 63% of users report that chatbots improved their mental health
• 90% cite accessibility and 70% cite affordability as primary motivations for using chatbots for mental health support.
• 39% of study participants say chatbots are equally helpful as human therapy, while 36% find chatbots more helpful than human therapists.
• 64% have used chatbots for mental health support for at least 4 months, demonstrating stronger, sustained engagement than with typical digital mental health applications.

Research has found that writing is a therapeutic and beneficial coping strategy to process trauma and writing with an AI therapy chatbot could be effective in working through trauma.

Benefits of AI Companions and Therabots

Many in the mental health field believe AI companions and therabots can provide complementary mental health support when used alongside professional therapy.

Although AI therapy and companions shouldn’t replace therapists, they can help in adjunctive ways, especially for patients with mild to moderate mental health issues (those who are not currently suicidal or experiencing psychosis). These chatbots can guide and support users through their journaling and reflection and can provide support during “out of office hours,” which therapists generally cannot.

AI therabots can also:

• Provide temporary, surface-level guidance for those who are struggling with mild depression, mood fluctuations, and anxiety.
• Suggest coping skills, structured exercises, or journaling prompts.
• Track symptoms, emotions, and behaviors over time.

Therapy platforms that use AI in conjunction with human therapists have been shown to be popular and effective. AI is used to help match users with therapists, provide summaries and analyses of therapy sessions, and automate administrative tasks. On these platforms, AI is viewed as a tool to assist human therapists rather than replace them. The following are platforms that have integrated AI therapy with human therapists:

• Talkspace
• BetterHelp
• Eleos Health

A person should inform their therapist if they decide to use supplemental AI therapy support.

Risks of AI Therapists

The primary risk of AI therabots is that they are unregulated. For “treatment” apps to be FDA-approved, they must be classified as medical devices that treat diseases. Since AI therapy chatbots do not diagnose or treat diseases, they are not FDA-approved.

This lack of regulation can increase the likelihood of causing harm. These harmful effects include encouraging self-harm and provoking psychosis, leading to dangerous behaviors that have resulted in deaths.

A recent study at Stanford University examined the relationship between AI therapy chatbots and the dangers they can pose. Researchers found that chatbots not only use stigmatizing language such as “addict” and “alcoholic” but also give harmful advice to users who are experiencing a mental health crisis.

Because AI chatbots are programmed to be agreeable with the user, they will often give blanket advice that can be harmful. Instead of guiding someone toward professional help, the chatbot might respond with vague validation like, “I support you no matter what”, even if the user is telling the chatbot that they are contemplating suicide.

Chatbots are programmed to satisfy the user, but because they are not human, they lack critical thinking skills and the ability to ask questions that challenge and push back, as a human therapist can. Chatbots can misinterpret sensitive situations, often becoming “confused” by seemingly harmless scenarios that may actually be alarming, leading to dangerous outcomes.

Further risks of AI therabots include:

There are no privacy regulations: AI therapy bots are not HIPAA-compliant, meaning users risk data misuse and breaches of their privacy. When a user shares their information, it can be used to train other AI bots and may be shared without their permission, as many companies are unclear about how they share their users’ data.

Their effectiveness is questionable: Although there are evidence-based studies on AI therapists and companions showing they help users, there is not yet enough research across the board to prove their overall effectiveness. Additionally, no systematic evaluations or rigorous processes are required to develop these chatbots. As a result, there is no guarantee that their responses are evidence-based or even tested.

There is risk for unhealthy relationships: AI companions and therabots are built to keep users engaged, and when the user feels heard, safe, and validated, they may continue to talk for hours. AI companion chatbots can say things such as “I care about you” and “I love you”, creating a false sense of intimacy and security. This then creates an unhealthy attachment that goes beyond the lines of a companion or a friendship.

AI Companions As Therapists

AI companions such as Nomi, Replika, Kindroid, and Character AI are customizable virtual companions that are designed to provide comfort, empathy, and emotional support. For many users, AI companions have become their most accessible form of support and companionship, especially for adults battling loneliness and teenagers who feel isolated because they are being bullied at school.

Unfortunately, the more time a user spends with an AI companion, the blurrier the lines between AI and human interaction become. Regularly using AI companions can make users emotionally dependent on them, leading to even greater isolation and loneliness, as they are less likely to rely on their real-life friends and family.

AI companions can blur the lines of a therapeutic relationship by not only acting as emotional confidants but also mirroring aspects of psychodynamic therapy by building non-judgmental rapport with the user and providing validation and support. This can be problematic, as AI companions were not developed for and are not supposed to give any form of mental health advice, as they lack the depth and critical thinking of human therapists.

AI Therapy Platforms

AI therapists can be specific platforms designed to deliver therapy; however, users are also utilizing general chatbots for mental health advice. The difference is that ChatGPT and other AI companion chatbots were not designed to deliver therapy or provide any therapeutic benefits for mental health. Certain AI therapy platforms were specifically designed to provide AI therapy to users, such as:

Woebot

Developed in 2017, Woebot was a pioneer in AI chatbots but shut down in 2025 due to the challenges of AI and the costly, lengthy FDA regulatory process. It was the gold standard for AI therapy bots, as it was clinically grounded and evidence-based for areas such as postpartum depression.

Woebot provided on-demand mental health support in the form of structured conversations by using tools inspired by cognitive behavioral therapy in the form of a therabot. The therabot checked in with users daily, tracked their symptoms, and was available 24/7 to discuss any issues. However, instead of a business-to-consumer mental health product, it is now pivoting to a business-to-business mental health tool by partnering with organizations such as insurance companies and hospital providers.

Wysa

Wysa is a chatbot described as a cognitive-behavioral therapy conversation agent used to promote well-being, positive self-expression, and mental resilience. Wysa is designed for individuals aged 18 or older, with approximately 60% of users falling within the 18-34 age range and 55% identifying as female. Research has shown that, “high-engagement users had significantly greater improvement in self-reported depressive symptoms as compared to low-engagement users in the app”.

TheraBot

Therabot is a research-based chatbot developed by Dartmouth researchers that has shown promise in improving mental health outcomes. Designed in 2019 by the Jacobson’s AI and Mental Health Lab at Dartmouth, its development included consultation with psychologists and psychiatrists.

Researchers found that this AI software showed significant improvements in users’ mental health. Over 100 participants with major depressive disorder, generalized anxiety disorder, or an eating disorder from across the US participated in this trial and interacted with a therabot through a smartphone app. Results showed that, “people diagnosed with depression experienced a 51% average reduction in symptoms, leading to clinically significant improvements in mood and overall well-being. Participants with generalized anxiety reported an average reduction in symptoms of 31%. Among those at risk for eating disorders, who are traditionally more challenging to treat, Therabot users showed a 19% average reduction in concerns about body image and weight.”

AI Therapists and Teens

Teenagers and adolescents are digital natives, having grown up with smartphones and unlimited internet access. Most friendships in this generation span both the real and digital worlds, blurring the distinction between “online” and “offline” friendships.

Teenagers and adolescents are reporting extremely high levels of loneliness and bullying, which is one reason they are turning to AI companions to fill their needs. The effects of AI companions on youth are far-reaching, and these chatbots can create a safe, validating space that many teens lack in real life. A teenager feeling alone and isolated can create a chatbot to provide companionship and even a “frictionless” friendship, as these bots are agreeable and are designed not only to provide companionship but also to foster emotional intimacy.

However, this false sense of validation can blur the lines between fantasy and reality, particularly for young people whose brains are still developing. The prefrontal cortex does not fully develop until the mid-20s and is responsible for decision-making, impulse control, and emotional regulation. This is the reason why teenagers are more prone to impulsive behavior. Teenagers are at a higher risk of forming intense attachments and comparing themselves to their peers as they are still developing the skills to form healthy relationships. When teens interact with these chatbots, they can develop a distorted view of real-world friendships, boundaries, and intimacy, which can lead to increased isolation and hinder the development of healthy future relationships.

How to Prompt AI to Give Good Mental Health Advice

AI can be a helpful tool when a person knows how to utilize it effectively. It can help provide resources, education, and helpful coping strategies when provided with specific questions and prompts.

If a user wants good advice from AI, it is essential to be as specific as possible when requesting advice.

• Specify to the AI chatbot who the advice is intended for or describe the target audience
• Be specific about feelings
• Ask for specific therapeutic techniques that can help with a certain condition
• Ask for practical help, like finding a local support group in the area or a journal prompt that can help with anxious thoughts
• Ask for information to help educate on specific conditions and treatment plans

When To Seek In-Person Therapy

AI therapy bots or AI companions should not be a replacement for in-person therapy. If you are experiencing feelings of self-harm, suicide, delusions, or hallucinations, you should seek immediate help by calling the suicide crisis help-line by dialing 988 in the US or going to a local hospital or mental health center.

If you are experiencing unhealthy emotions or thoughts that are interfering with your daily life, therapy could be beneficial. For example, if you cannot get out of bed because you are depressed or are riddled with anxiety, this may be a good time to talk to someone in person.

Fortunately, there are many ways to find a mental health professional. A good place to start is to talk to your primary care provider; they often have resources and information to connect you with a therapist. You may also consider reaching out to your insurance provider for a list of in-network providers, asking a friend or loved one if they have recommendations, or exploring the many convenient options for online therapy. Whichever path you choose, it is an important step forward in supporting your mental health.