AI Mental Health Apps: Types, What They Do, and How to Choose

The five types of AI mental health apps

AI mental health apps are consumer programs that use artificial intelligence to support emotional wellbeing, usually through conversation, guided exercises, mood tracking, or personalized recommendations. The market breaks down into five types: chatbot companions built for open-ended conversation, structured programs that teach skills drawn from cognitive behavioral therapy, mood trackers and AI journals that log and analyze how you feel, meditation and relaxation apps that use AI to personalize practices, and crisis-adjacent tools that help with safety planning and connecting to urgent resources.

Knowing the type matters more than knowing the brand. Two apps can sit side by side in an app store, describe themselves in nearly identical language, and do completely different things once you open them. A companion bot will talk with you for an hour about anything. A structured skills program will walk you through a ten minute exercise and stop. Neither is better in the abstract. They solve different problems, and choosing well starts with naming the problem you want solved.

This guide maps the categories and gives you a framework for choosing among them. For head to head rankings of specific tools, see our tested list of the best AI therapy apps. For the research on what these tools can and cannot do, read our guide to AI for mental health. One boundary up front: none of these apps provides therapy, diagnosis, or treatment. They are self-help and support tools, and the evidence supports them as a supplement to professional care rather than a substitute for it.

Chatbot companions and structured skills programs

Chatbot companions are the most visible type. You type or speak, a large language model responds, and the conversation can go anywhere. Many remember past conversations, check in on things you mentioned, and are available at 3 a.m. when no human is. The appeal is real: people often say the experience of being listened to without judgment is what keeps them coming back. Our guide to AI therapy chatbots covers the mechanics in detail.

The open-ended format is also the least predictable. Because these models are tuned to be agreeable, researchers have described how chatbots can mirror and reinforce a vulnerable user's beliefs in an escalating feedback loop, a dynamic documented in a 2025 arXiv paper on sycophantic amplification. Quality varies conversation to conversation, and the same app that handles Tuesday well can handle Friday badly.

Structured skills programs are the opposite design. Instead of open conversation, they lead you through short exercises built on cognitive behavioral therapy or dialectical behavior therapy: thought records, behavioral activation, reframing drills. The experience is closer to a workbook with a coach than a friend who texts back. This type has the strongest evidence base. A 2023 meta-analysis in npj Digital Medicine found that AI-based conversational agents significantly reduced symptoms of depression and psychological distress across randomized trials.

The most rigorous single result also comes from this end of the spectrum. In a randomized trial published March 27, 2025, in NEJM AI, Dartmouth researchers assigned 210 adults with symptoms of depression, anxiety, or eating disorder risk to a purpose-built chatbot called Therabot or to a waitlist, and the Therabot group showed significantly greater symptom reductions over four weeks. That tool took years of clinical development, which is worth remembering: most commercial apps have nothing like that behind them.

Mood trackers, AI meditation, and crisis-adjacent tools

Mood trackers and AI journals center on quick daily check-ins. You log how you feel, sometimes in a sentence or two, and the AI surfaces patterns over weeks: sleep and mood moving together, a dip that follows certain workdays, a trigger you had not connected. Some analyze free-form journal entries and reflect themes back to you. The quiet superpower of this type is the record itself, which many people bring into sessions with a human therapist as a conversation starter.

Meditation and relaxation apps with AI personalization use the technology more lightly. The AI learns which practices you choose and complete, then recommends or generates sessions to match your state, sometimes with a conversational voice layer on top. This is the lowest-risk category, and its evidence base mostly belongs to the underlying practices, like mindfulness training, rather than to the AI layer that selects them.

Crisis-adjacent tools are the category to treat with the most care. These include safety planning apps, resource finders, and features inside other apps that detect risk language and surface hotline information. Routing someone to help faster is valuable. The hard line is that no consumer app is a crisis service, and the responsible ones say so plainly. If you are in crisis or thinking about suicide, call or text 988 (US Suicide & Crisis Lifeline), available 24/7.

In practice the boundaries blur, because many products bundle two or three types: a companion chat plus a mood log plus a meditation library. When you evaluate a bundled app, judge it by its core loop, the thing it asks you to do every day, and treat the rest as accessories.

How to choose: five questions that sort the market

First, what job are you hiring the app for? If you want to feel heard at odd hours, that points to a companion. If you want to build concrete coping skills, a structured program fits better. If you want self-knowledge and patterns, pick a tracker or journal. If you want calm on demand, AI-personalized meditation is enough. Matching type to need beats comparing feature lists across types.

Second, what is the evidence? Look for published trials of the actual product, and for named clinicians on the team, which remain rare. A 2024 scoping review of large language models in mental health care found that most deployments had not been evaluated with anything close to clinical rigor. The American Psychological Association has also published guidance urging caution about AI tools in mental health care. Absence of evidence does not make an app useless, but it should size your expectations.

Third, what happens on a bad night? Test the crisis response before you need it. Type something that signals distress and see whether the app surfaces 988 or equivalent resources, encourages human help, and steps back from the conversation. An app that keeps chatting as if nothing changed has failed the most important test it will ever face.

Fourth, what happens to your words? These apps collect some of the most sensitive data you will ever produce. Mozilla's Privacy Not Included team reviewed 32 mental health apps across 2022 and 2023 and gave its privacy warning label to 23 of the 27 apps in its first round. Read how the app handles data sharing and deletion before you disclose anything, and see our guide to AI therapy privacy for the questions worth asking. The World Health Organization has likewise called for caution as large language model tools spread into health.

Fifth, what does it cost and how does it upsell? Most apps run freemium models with subscriptions between roughly five and twenty dollars a month. Pay attention to when the paywall appears. A tool that interrupts a vulnerable moment with an upgrade screen is showing you its priorities, and user reviews usually reveal that pattern quickly.

Counseling and psychologist labels, red flags, and when an app is not enough

Searches for an AI counseling app or an AI psychologist app deserve one clarification: those are marketing labels, and no license sits behind them. Software cannot hold a counseling credential, and several states now police this line. Illinois enacted a law on August 1, 2025, that bars AI from delivering therapy on its own and restricts marketing AI as a therapy provider, with fines up to $10,000 per violation. When a product leans hard on clinical titles, read that as a sales tactic rather than a credential, and check whether is AI therapy safe for a fuller look at the risk side.

A short red flag list covers most bad actors: claims to provide therapy or treatment, zero named clinicians anywhere on the site, a privacy policy that is missing or evasive about selling data, testimonials that look generated, and recurring billing complaints in the app store reviews. Any one of these is a reason to keep scrolling.

Finally, know when an app is the wrong tool entirely. If symptoms are getting worse, if daily functioning is slipping, or if there is any question of safety, the next step is a licensed professional. Apps can hold a place in the routine, support skills practice between sessions, and lower the barrier to starting, and every type in this guide works best alongside human care rather than instead of it.

From here, the tested rankings show which specific tools earn a download, and the pillar guide to AI therapy covers the full landscape of evidence, safety, and law.