Christy Chen on why today’s chatbots aren’t built to help you know yourself

Cit Labs founder Christy Chen argues that current AI chatbots are poorly suited to emotional support because they are optimized for engagement, trained on performative support text, and lack physiological inputs. The startup is developing an alternative using EEG data and clinician-defined guardrails, positioning its product as a purpose-built mental-health-support technology. The article presents an early-stage product thesis rather than material financial or market-moving developments.

Analysis

This is not yet a monetizable public-markets catalyst: the proposed differentiation—proprietary physiological data plus clinician-designed safety constraints—implies a materially slower, more regulated development path than consumer chatbot deployment. The economic moat, if validated, would reside in exclusive longitudinal biosignal datasets and clinical evidence rather than model quality; that favors specialist health-data platforms over general-purpose LLM vendors, but no listed beneficiary is sufficiently direct today.

The second-order implication is that mental-health AI may bifurcate into low-cost consumer engagement products and higher-cost, clinically supervised decision-support tools. The latter category faces reimbursement, privacy, medical-device classification, and malpractice-liability gating over 6-18 months, limiting near-term revenue expectations for the broader AI-healthcare complex. A safety incident involving a consumer AI companion could accelerate demand for clinically governed products, but would more immediately create regulatory and multiple-compression risk for consumer-facing AI names with aggressive healthcare-adjacent narratives.

Contrarian view: the market may overestimate the value of adding hardware-derived signals before proving that they improve clinical outcomes versus structured patient-reported data. EEG collection adds friction, device cost, and potentially weakens retention—the core consumer-AI economic engine. The relevant falsification metric is not engagement; it is independently replicated improvement in clinical outcomes, crisis-escalation accuracy, and reimbursement coverage versus conventional teletherapy workflows.

Key Decisions for Investors

• No directional trade from this item. Treat private neurotechnology and clinical-AI claims as diligence alerts rather than read-throughs to mega-cap AI valuations.
• Monitor HIMS and TDOC over the next 1-3 months for any expansion of AI-led mental-health offerings. A move toward autonomous care without disclosed clinician oversight, adverse-event reporting, or payer reimbursement should be viewed as a downside-risk signal rather than a near-term margin catalyst.
• Maintain a watchlist on NPCE as the closest listed neuro-data proxy, but do not initiate solely on this theme; require evidence of non-epilepsy data monetization, a commercial partnership, or regulatory clarity before assigning incremental value.
• For broader healthcare-AI exposure, favor companies with established clinical workflow distribution over consumer-chatbot narratives; reassess if FDA guidance or a major payer creates a defined reimbursement pathway for AI-enabled behavioral-health monitoring.