The GPT Oracle: 1 in 10 Americans are reviving the pagan practice of prophecy

A 2026 Pew report found that 10% of Americans use chatbots for emotional support or companionship, while an analysis of 1.1 million ChatGPT conversations found 49% involved users seeking information or advice. The article argues that AI systems' opacity may reinforce their perceived authority, drawing parallels with historical oracles. It highlights AI-adoption and trust dynamics rather than a material near-term financial catalyst.

Analysis

This is not a near-term earnings catalyst, but it reinforces that consumer AI value is migrating from search and task completion toward recurring, high-trust engagement. The strategic beneficiaries are the distribution owners—META, GOOGL, MSFT and AMZN—because emotionally salient use cases raise switching costs, interaction frequency and the amount of first-party preference data available for personalization. The monetization lag is material: higher engagement does not translate into revenue unless platforms can introduce paid tiers, commerce referrals or advertising without damaging perceived neutrality.

The underappreciated risk is not merely model hallucination; it is liability and reputational asymmetry. A highly publicized harmful recommendation in mental-health, medical, financial or self-harm contexts could trigger app-store restrictions, state attorney-general action and product-design changes that impair engagement precisely in the highest-retention use cases. Over the next 1-3 months, this is principally a regulatory-headline and sentiment variable; over 6-18 months, it favors firms with distribution, moderation infrastructure and balance sheets sufficient to absorb safety/compliance costs, while pressuring smaller consumer-AI applications that lack proprietary distribution or a clear liability framework.

Consensus may overstate the immediate revenue opportunity from companionship and advice while understating the eventual cost of trust-and-safety. Consumer willingness to consult a model is not equivalent to willingness to pay, and introducing ads or affiliate economics into advice-oriented conversations could create an adverse-selection problem: the most commercially valuable recommendations are also the most scrutinized. The key falsifier of the cautious view would be evidence that paid AI subscriptions retain users at scale without elevated safety incidents or that major platforms establish durable, regulator-accepted guardrails for high-stakes queries.

Key Decisions for Investors

• No standalone directional trade from this item; treat it as a watch signal rather than an earnings catalyst given the absence of measurable incremental usage, conversion or ARPU data.
• Maintain a 6-18 month quality bias toward GOOGL and MSFT versus subscale consumer-AI software names: both can bundle AI into existing enterprise/consumer distribution and fund compliance, while smaller applications face higher customer-acquisition and liability costs. Reassess if paid-assistant retention data show independent apps sustaining materially lower churn.
• Monitor META for a more immediate engagement upside, but do not add solely on AI-companionship narratives; require disclosure of incremental messaging engagement or monetization. A material safety controversy involving AI interactions would be a catalyst to reduce exposure, as reputational risk is amplified in social products.
• Set alerts for US/EU rules specifically covering chatbot duty of care, minors, mental-health claims or recommendation disclosures. Such rules would likely widen the moat for MSFT, GOOGL and AMZN while compressing valuation multiples for unprofitable consumer-AI peers.