The Short Answer

Some AI therapy apps have demonstrated statistically significant symptom reduction in peer-reviewed studies. Wysa and Woebot (now enterprise-only) have the strongest evidence. But "effective" must be qualified: effective for whom, for what severity, and compared to what alternative. Most studies show benefit for mild-to-moderate symptoms. No study has demonstrated AI therapy equivalence with face-to-face therapy for serious mental health conditions.

What the Research Shows: Platform by Platform

Wysa — Strongest Available Evidence (30+ Papers)

Wysa has the most extensive research base of any currently available consumer AI therapy app. Its published evidence includes randomized controlled trials (RCTs) in JMIR demonstrating significant symptom reduction for anxiety and depression. Studies have also validated its use for chronic pain management, sleep difficulties, and workplace stress.

A particularly noteworthy finding: research on therapeutic alliance showed users formed bonding levels with the Wysa chatbot comparable to human therapists. This challenges the assumption that AI cannot create a meaningful therapeutic relationship.

Regulatory validation: Wysa is CE-marked as a Class I medical device in the EU and endorsed by the UK NHS through its Innovation Accelerator programme.

Woebot — Most RCTs, But No Longer Available to Consumers

Woebot had 14+ randomized controlled trials — the most of any AI therapy chatbot globally. Founded by Stanford clinical psychologist Dr. Alison Darcy, it was built on rigorous CBT, IPT, and DBT foundations. However, Woebot permanently shut down its consumer app on June 30, 2025, citing FDA regulatory burden. It now operates exclusively through enterprise health system partnerships.

This creates a paradox: the most clinically validated AI therapy chatbot is no longer accessible to the consumers it was studied on. Woebot's research remains valuable for understanding AI therapy effectiveness generally, even though the specific product is unavailable.

Youper — Promising Observational Data

Youper has one major published study — a Stanford University observational study with 4,517 participants showing significant reductions in anxiety (effect size d=0.57) and depression (d=0.46) within 2 weeks. The study also reported 42.66% 4-week retention, which is high for a mental health app.

Important caveat: This was an observational study without a control group. Participants self-selected, and there's no way to determine how much improvement would have occurred without the app. An RCT is needed to establish causation.

Elomia — Early Clinical Trials

Elomia has active clinical trials (NCT06725147) and a quasi-experimental study published in BMC Psychology showing significant improvements in somatic symptoms, anxiety/insomnia, social dysfunction, and depression among college students. The research base is growing but not yet at the level of Wysa's.

Replika, MindDoc, Talkiatry, Bloom

Replika has mixed evidence — some studies show loneliness alleviation, but others document harms from emotional dependence. It is explicitly not a clinical tool. MindDoc uses validated assessment instruments (PHQ-9, GAD-7) but has no published RCTs about its overall platform effectiveness. Talkiatry reports internal outcomes data (not peer-reviewed). Bloom has no independent clinical studies.

What "Effective" Actually Means

Research on AI therapy effectiveness must be interpreted carefully:

Effect sizes are modest. Youper's d=0.57 for anxiety is a "medium" effect — meaningful but not transformative. Compare to face-to-face CBT, which typically shows effect sizes of d=0.8-1.0.
Mild-to-moderate conditions. Studies consistently show benefit for mild-to-moderate anxiety and depression. Evidence for severe conditions, psychosis, PTSD, personality disorders, and complex trauma is absent or negative.
Self-selected participants. People who download and use therapy apps are not representative of all people with mental health conditions. Study results may not generalize.
Short follow-up periods. Most studies measure outcomes at 2-8 weeks. Long-term effectiveness data is limited.
Supplement, not replacement. Research positions AI therapy as a supplement to professional care or a bridge for people who cannot access therapy — not a replacement for treatment.

When AI Therapy Apps Are Appropriate

Mild-to-moderate anxiety or depression symptoms
Stress management and emotional regulation skill-building
Between-session practice for people already in therapy
First step for people not ready for or unable to access traditional therapy
Sleep support and mindfulness practice
Psychoeducation about therapy modalities (CBT, DBT, ACT)

When AI Therapy Apps Are NOT Appropriate

Active suicidal ideation or self-harm — call 988 immediately
Severe depression, psychosis, or mania
Active substance use disorders
Complex PTSD or trauma processing
Personality disorders requiring specialized treatment
Any condition requiring medication management
Children and adolescents (most apps are validated only for adults)

The Bottom Line

AI therapy apps with clinical evidence (primarily Wysa) can provide meaningful symptom relief for mild-to-moderate anxiety and depression. They are most useful as supplements to professional care or bridges for people facing access barriers. They are not replacements for therapy with a licensed professional, especially for serious mental health conditions.

The critical consumer action: check whether the app you're considering has published peer-reviewed research. Most do not. Among currently available consumer apps, Wysa (30+ papers) has a dramatically stronger evidence base than any competitor. See our Clinical Evidence Scorecard for a complete breakdown.

Is AI Therapy Effective? What the Research Actually Says