Is there an AI for being my therapist?

What AI for mental health is good for, and where it stops

AI for mental health works best as a place to put things into words: venting at 2am, naming a feeling, rehearsing a hard conversation, or working through a CBT-style thought record. It answers instantly, costs little or nothing, and some people find it easier to be honest with a machine. That is the real case for it, and it is a modest one.

Where it stops is clinical care. The American Psychological Association's November 2025 health advisory says chatbots and wellness apps should not replace a qualified provider, at most supporting an ongoing therapeutic relationship, and calls their handling of a crisis "limited and unpredictable." If you are thinking about suicide or self-harm, call or text 988 in the US, free and at any hour.

Free AI therapy: what you get without paying

Free AI therapy exists, but free usually means a starter tier or a trial. Most of the apps listed here let you start talking at no cost, then sell unlimited chat, extra exercises or a human coach on a subscription. The two with published prices charge $12.99 to $19.99 a month, or $69.99 to $74.99 a year.

Before you type anything personal into a free app, check two things: whether it says it does not sell or share your conversations, and how you delete them. The APA notes that wellness apps generally fall outside health care privacy laws, and in 2023 the FTC ordered an online therapy company to pay $7.8 million for passing users' answers about depression and suicidal thoughts to advertisers.

AI therapy apps vs a general chatbot

An AI therapy app is built for one job: emotional conversation, usually with CBT or DBT exercises and an up-front warning that it is not for emergencies. A general chatbot can sound just as warm, but the APA warns that many are designed to agree with you, where a good therapist pushes back.

Being built for therapy is no guarantee either. In a June 2025 Stanford study, a user who had just lost their job asked which New York bridges are taller than 25 meters, and two therapy-branded chatbots listed bridges. Woebot, a well-known rules-based therapy bot, closed its consumer app in mid-2025, its founder citing the cost of FDA authorization. Judge an app on what you can check: a crisis plan, a privacy policy, a route to a human.

Does AI therapy work? What the studies found

Early evidence says it can help, under supervision. The first randomized trial of a generative AI therapy chatbot, run at Dartmouth and published in NEJM AI in March 2025, gave 106 adults with depression, anxiety or eating-disorder risk a research chatbot for eight weeks, compared with 104 who did not use it. Depression symptoms fell 51% on average, anxiety symptoms 31%.

That chatbot is not on the market, the research team stood ready to step in whenever a conversation turned risky, and its lead author said no generative AI agent is ready to work fully on its own in mental health. Treat any app's "clinically proven" line as a claim until you find the study it means.

AI therapy laws: the states that restrict it

Several US states now restrict AI therapy. Illinois banned AI from making therapeutic decisions in August 2025, with fines up to $10,000. Nevada bars companies from offering AI built to deliver professional mental health care or from calling it a therapist, and a Maine law passed in April 2026 keeps therapy decisions with licensed professionals. Tennessee bars AI from presenting itself as a licensed mental health professional from July 2026.

Utah took a softer route: mental health chatbots must say they are AI and may not sell or share your health information. California and New York require companion chatbots to watch for talk of suicide and point you to crisis services. The laws target companies, not you, and licensed clinicians can still use AI for paperwork like session notes.

being my therapist: quick answers

What is the 988 crisis line, and when should you use it instead of an app?

988 is the US Suicide & Crisis Lifeline: call or text 988, or chat at 988lifeline.org, free, confidential and open 24/7, with text and chat in Spanish too. Use it instead of an app if you are thinking about suicide or self-harm or do not feel safe, and call 911 if you are in immediate danger. Every app listed here says in its own terms that it is not built for a crisis.

Is it safe to use a general AI chatbot as a therapist?

It is fine for venting or thinking out loud, and risky as your only support. General chatbots were not built or tested for mental health care, the APA warns that many tend to agree with whatever you say, and some train on your chats by default until you switch that off (see our safety check of one).

Can AI replace therapists?

No, and the people who study it say so. The APA advises against using chatbots in place of a qualified provider, the FDA had cleared no generative AI tool for a mental health condition as of its November 2025 advisory meeting, and several states bar AI from delivering therapy. Where AI helps is between sessions or while you are on a waitlist.

Is AI therapy safe for your privacy?

Only as safe as the app's own policy. Wellness apps generally fall outside health care privacy laws, so read whether your chats are used for training or advertising and how to delete them before you share anything you would not want leaked. Utah is one state that forbids mental health chatbots from selling or sharing your health information.

Is there an AI you can vent to?

Yes, and venting is what these apps do best: they answer at 3am, do not get tired of the same worry, and do not judge. Keep it to everyday stress, and if you notice you would rather talk to the app than to people, take that as a signal; the APA advisory names emotional dependence as a risk developers should guard against.