Amid a shortage of therapists, experts discuss how AI can make mental health treatment cheaper and accessible, its privacy concerns, product efficacy, and more
Context & Ripple Effects
This WSJ roundtable lands at the start of a five-year arc the corpus now traces end to end: in 2021, experts frame AI as a supply-side fix for a therapist shortage while flagging privacy and unproven efficacy. Within days of publication, virtual therapy apps like Talkspace and BetterHelp were being criticized for promising more than they deliver, and by late 2022 chatbots such as Wysa and Woebot had millions of users despite scant research supporting their efficacy.
What makes the early framing worth revisiting is how literally its warnings played out: by 2024 users were choosing chatbot therapists precisely because they are cheap, quick, and available around the clock, and by 2026 eating disorder therapists report patients using AI chatbots in ways that undermine clinical advice. The shortage argument won; the safeguards did not arrive first.
First-order effects
- Patients facing long waits and high costs gain an immediate alternative — cheap, 24/7 conversational support — which shifts demand away from the scarce human therapists this coverage says are in short supply.
- App operators like Talkspace and BetterHelp ride the same demand surge, and face immediate scrutiny that their products overpromise relative to what they can deliver.
Second-order effects
- Chatbot makers such as Wysa and Woebot absorb the pressure of the efficacy critique, competing on user scale even as researchers note little evidence behind the tools.
- Human practitioners begin contending with AI as a counterparty rather than just a tool: eating disorder therapists report chatbot use actively undermining their treatment plans, while helplines find AI routing more patients to them.
Third-order effects
- If adoption keeps outpacing evidence, mental health care structurally splits into two tiers — clinician-led treatment and ungoverned consumer chatbots — forcing regulators to decide where privacy protections and efficacy standards apply to conversational AI.
- The pattern points toward formal governance of AI companions in sensitive domains, because the harm cases emerging in therapy are the template for how companion-style AI gets regulated everywhere else.
The trend: Mental health delivery is migrating from scarce human clinicians toward always-available AI support, with efficacy research and privacy governance consistently trailing adoption.