Millions of Americans Are Talking to AI Instead of Going to the Doctor, and It's Giving Them Horrendously Flawed Medical Advice

A JAMA Network Open study found frontier LLMs failed to provide accurate medical advice more than 80% of the time in ambiguous cases and 40% of the time even with clearer symptoms, exam findings, and lab results. The article also cites a West Health-Gallup survey showing 25% of U.S. adults, or about 66 million people, are asking chatbots for medical advice, with 14% of those users saying they never saw a provider they otherwise would have. The piece underscores rising patient reliance on AI and the growing case for regulatory oversight, with potential implications for AI and healthcare adoption.

Analysis

GOOGL is the obvious first-order loser, but the more important read-through is to the entire consumer-facing AI stack: this increases the probability that generic, unspecialized copilots get pushed out of high-stakes workflows in favor of vertically constrained, auditable systems. That favors companies with distribution inside regulated clinical channels, structured data moats, and liability-aware product design; it is a headwind for broad “AI assistant” monetization narratives that rely on consumer trust rather than verified task completion.

The second-order effect is that scrutiny shifts from model quality to deployment architecture. In healthcare, value accrues less to the model vendor and more to the control layer: logging, provenance, human-in-the-loop escalation, and EMR integration. That creates a relative advantage for incumbents with workflow ownership and a relative disadvantage for cloud/model providers whose products can be commoditized or excluded from sensitive use cases by policy rather than performance.

For GOOGL specifically, the risk is not near-term ad revenue from AI Overviews alone, but a slower erosion of trust that widens the legal/regulatory discount on AI surface area across Search, Gemini, and healthcare-adjacent products. Time horizon matters: over days this is likely a sentiment drag; over months it can become a product liability and monetization headwind if regulators or health systems begin formal procurement restrictions. The upside reversal case is narrow: demonstrable gains in calibrated refusal behavior, medical-tuned models, or third-party validation that materially improves precision on ambiguous cases.

The contrarian view is that consensus may be overestimating immediate revenue damage while underestimating the strategic importance of being present in the clinical workflow. If AI becomes a front-end triage layer with guardrails, the winners may still be the platform owners that can route demand, even if they do not fully solve diagnosis. The market may be pricing this as a pure brand issue; the deeper issue is whether GOOGL can monetize trusted, bounded AI without inviting a regulatory overhang that slows experimentation.

Key Decisions for Investors

• Reduce near-term GOOGL exposure; if holding, hedge with 1-3 month put spreads to capture potential regulatory/sentiment drift while limiting premium bleed.
• Long pair: TDOC or HIMS vs short GOOGL over 2-6 months if you want a cleaner exposure to consumer health decision-making shifting toward trusted healthcare interfaces rather than general-purpose AI.
• Watch MSFT and AMZN as relative beneficiaries if enterprise/healthcare buyers favor auditability and workflow control; use any broad AI selloff to add on weakness versus GOOGL.
• For event-driven traders, fade a sharp post-headline bounce in GOOGL only if implied volatility remains elevated; the better risk/reward is a defined-risk put spread, not outright short stock.
• Selective long in healthcare IT vendors with EMR/workflow integration over 3-9 months, as health systems are likely to budget for governance, provenance, and escalation tools rather than generic chatbot access.