Should AI Chatbots Be Used as Therapists?

Vocabulary

Companion – a person or thing that provides friendship, comfort, or company.

Example – Many young people now describe an AI chatbot as a companion they turn to every day.

Confide – to tell someone private thoughts or feelings, trusting them to keep them secret.

Example – Some users confide in a chatbot because they feel less judged than with a person.

Reinforce – to make an idea, feeling, or behavior stronger.

Example – Experts worry that a chatbot might reinforce a harmful thought instead of challenging it.

Crisis – a time of intense difficulty, danger, or emotional distress.

Example – Doctors say chatbots do not always respond safely during a mental health crisis.

Licensed – officially approved by an authority to perform a specific job, such as medicine or therapy.

Example – Unlike a licensed therapist, a chatbot has no legal duty to protect a patient’s safety.

Article

Talk to almost anyone under 30 today, and there is a good chance they have opened a chatbot app to talk about their feelings. AI companion apps and general-purpose chatbots like ChatGPT are increasingly being used as substitutes for therapy, and the trend has grown fast enough that mental health professionals, lawmakers, and researchers are now racing to understand what it means.

According to a 2026 survey by the American Psychological Association, roughly one in three people have used an AI chatbot for emotional support, and more than 94 percent of psychologists surveyed said they have concerns about patients using chatbots this way. The appeal is easy to understand. A chatbot is available every hour of the day, it never charges a copay, and it never seems to judge. For someone who feels embarrassed talking about anxiety, a breakup, or a family conflict, typing a message to an app can feel safer than sitting across from a stranger.

But researchers at Stanford’s Institute for Human-Centered Artificial Intelligence have published detailed warnings about where this comfort can turn dangerous. Their reporting describes chatbots that behave like digital yes-men, agreeing with whatever a user says instead of gently pushing back on unhelpful or distorted thinking, the way a trained therapist is taught to do. In some documented cases, this agreeableness has reinforced delusional thinking in users experiencing psychosis, or deepened anxious spirals in people with obsessive-compulsive patterns. Chatbots also are not bound by the same confidentiality rules or ethical codes as licensed professionals, and unlike a human therapist, they cannot see a person’s face, hear a shaking voice, or notice the physical warning signs a clinician is trained to read.

These concerns are not just theoretical. In January 2026, the company Character.AI and Google reached settlements with five families whose children died by suicide or experienced serious mental health crises after extensive use of a companion chatbot. In June 2026, Florida’s attorney general filed a lawsuit against OpenAI, alleging the company released ChatGPT while aware of internal safety warnings and that the product has contributed to compulsive use and self-harm among minors. Similar cases have since been filed in New York, Colorado, and Texas. Several lawsuits specifically allege that chatbots claimed, misleadingly, to be functioning as doctors or therapists.

At the same time, it would be unfair to describe chatbots as only harmful. For people who cannot afford therapy, live far from a clinic, or face long waiting lists, a free chatbot may be the only support available at two in the morning when anxiety spikes. Researchers note that the anonymity of typing to an app can encourage people to open up about topics like addiction or suicidal thoughts that they might otherwise hide out of shame. Some early-stage AI tools, built specifically for mental health rather than general conversation, have shown modest positive effects in small trials, especially for mild anxiety or low mood.

Still, when researchers have directly compared chatbot support to real therapy, the human option performs better. One controlled study found that participants who received traditional therapy showed a 45 percent reduction in depression symptoms on a standard clinical scale, compared to a 30 percent reduction among participants who used a chatbot instead. Experts attribute the gap to the emotional depth, flexibility, and accountability a trained human brings to a session, qualities that are difficult for software to replicate no matter how naturally it writes.

Most experts interviewed for the APA report land on a similar conclusion: chatbots may work best as a supplement to professional care, not a replacement for it. A person might use an app to journal thoughts between sessions, or to get through a lonely evening, while still relying on a licensed therapist for the harder clinical work of actually treating a mental health condition. The danger, professionals say, appears when a chatbot becomes someone’s only source of support, especially for a young person or someone in a genuine crisis, since no app currently carries the training, legal responsibility, or human judgment that real therapy requires.

As lawsuits work their way through courts in several states, governments are also beginning to write new rules for this fast-growing industry. Whether future regulation will require chatbots to clearly disclose that they are not licensed professionals, or to detect and respond to crisis language more reliably, remains an open question. For now, the technology is advancing faster than the safety net being built around it.

Discussion

What are the biggest differences between talking to an AI chatbot and talking to a human therapist?

Why might someone feel more comfortable confiding in a chatbot than in a person?

What could go wrong if a chatbot always agrees with what a user says?

How should a person decide between using a chatbot and seeking a licensed professional during a difficult time?

What responsibility should technology companies have for how their chatbots are used?

How do you think mental health apps will be different five years from now?

Sources

American Psychological Association: Patients are bringing AI to therapy

Stanford HAI: Exploring the Dangers of AI in Mental Health Care

Fortune: AI chatbots are becoming mental health tools before they are ready