Risks & Limitations of Consumer AI Therapy Apps
No consumer AI therapy app is a substitute for a licensed mental health professional. The
category has real, documented limitations that any reader should weigh before relying on these
tools — especially for moderate-to-severe symptoms, crisis situations, or use by minors. The
points below summarise the published evidence and regulatory actions to date.
1. Crisis handling is uneven and not designed for emergencies
Most consumer AI chatbots surface crisis resources (988, Samaritans, Lifeline) when they
detect risk language, but the detection is imperfect. Reviews of conversational AI in
mental-health contexts have found inconsistent escalation behaviour, occasional dismissive
responses, and gaps when users describe risk indirectly. If you or someone you know is in
crisis, call 988 in the US, Samaritans 116 123 in the UK/IE,
Lifeline 13 11 14 in Australia, or local emergency services — not
an app.
2. Companion-style chatbots have faced regulatory scrutiny
Replika has been the subject of FTC complaints alleging deceptive marketing to vulnerable
users, and a separate civil lawsuit against Character.AI involves the death by suicide of
14-year-old Sewell Setzer III in 2024, after extended conversations with a Character.AI
persona. Companion-style chatbots are not clinical tools, do not have RCT evidence for
therapy outcomes, and should not be relied on as a substitute for professional care. We rank Replika
low and explicitly do not recommend it as therapy.
3. Effectiveness evidence is strongest for mild-to-moderate symptoms
Wysa, Woebot (now consumer-discontinued), and Youper (shut down September 2026) have published
peer-reviewed studies showing meaningful symptom reduction for mild-to-moderate anxiety and
depression. Evidence for severe depression, acute PTSD, psychosis, eating disorders, bipolar
disorder, and substance-use disorders is much thinner or absent. For these conditions, the
right intervention is a licensed clinician — not a chatbot.
4. Use by minors carries additional risks
The FTC has explicitly flagged AI chatbot safety for children and teens as an enforcement
priority. Most consumer AI therapy apps require users to be 18+ (or 13+ with parental
consent), but age-gating is rarely enforced. We do not recommend any of the apps on this page
as a primary mental-health support for under-18s without licensed clinical oversight. Parents
should review the privacy policy and crisis-response behaviour before allowing minors to use
any of these apps.
5. Data privacy and HIPAA
Most consumer AI therapy apps are not HIPAA-covered entities because they are direct-to-consumer
products, not health-care providers. Conversations with these chatbots may be used to train models,
sold to third parties, or stored indefinitely. Wysa and a handful of others publish clearer data-handling
policies and undergo SOC 2 audits; many competitors do not. Read each app's privacy policy
before sharing anything sensitive, and assume your data is not protected the same way it would be
in a therapist's office.
6. AI therapy works best as a supplement, not a replacement
The published evidence consistently positions AI therapy chatbots as a useful supplement to
professional care — for between-session skill practice, mood tracking, or
psychoeducation — not as a replacement. If you are weighing AI therapy as an alternative
to seeing a therapist, please read our guide on whether AI therapy is effective and our breakdown of AI therapy versus general-purpose chatbots like ChatGPT before deciding.
This section is informational and does not constitute medical or mental-health advice. If you
are in crisis, please call your local emergency services or a crisis hotline. See our editorial guidelines and fact-checking policy for how we evaluate
evidence.