Parks Associates: 63% of US internet households use generative AI

AI-tool usage reached 63% of U.S. internet households in 2026, up from 51% in 2025, while paid adoption reached 22%. Paid users reported materially higher NPS across major generative-AI platforms, led by DeepSeek at 59, Grammarly at 55 and Perplexity at 51, versus free-user scores of 15 and 10 for Grammarly and Perplexity. The research indicates consumers will pay for clear productivity or specialized value, although trust and safety concerns remain a constraint for AI use in higher-risk decisions.

Analysis

The survey supports a key monetization distinction: willingness to pay appears tied to workflow-specific utility rather than general-purpose consumer engagement. That favors enterprise-adjacent platforms with clear ROI and distribution into existing software budgets—MSFT, CRM, ADBE, and GOOGL—over consumer assistants that must overcome a zero-price anchor. NPS is a retention proxy, not proof of conversion: paid cohorts are self-selected power users, so extrapolating their satisfaction into broad consumer ARPU is likely premature.

For META, weak satisfaction among unpaid assistant users is more relevant to product-market fit than near-term earnings, because its AI payoff is principally engagement, ad ranking, creator tooling, and business messaging rather than a standalone subscription. The risk is that elevated AI capex is being capitalized by investors as a monetization option before measurable incremental ad pricing or click-to-message conversion emerges. Over the next 1-3 months, consumer survey data should have little bearing on estimates; the relevant catalyst is Meta's next disclosure on inference costs, ad-conversion lift, and WhatsApp business-agent adoption.

Contrarianly, the market may be overemphasizing consumer subscription penetration as the AI profit pool. The larger 6-18 month value transfer could accrue to companies that embed AI in paid seats or high-intent transactions, while free assistants become costly acquisition funnels. Safety concerns create a secondary advantage for incumbents with enterprise governance, auditability, and indemnification, but any high-profile consumer harm event would raise compliance costs and favor MSFT/GOOGL over smaller model providers.

Key Decisions for Investors

• Maintain META as neutral/underweight versus GOOGL over the next 1-3 months: pair long GOOGL / short META in equal beta, targeting a 5-8% relative move if GOOGL demonstrates AI-driven search or Workspace monetization while META's AI remains a cost center. Stop if META reports sustained ad-pricing acceleration attributable to AI or material WhatsApp agent revenue.
• Accumulate MSFT on pullbacks for a 6-18 month horizon rather than chase consumer-AI headlines; enterprise distribution and paid-seat packaging make monetization more measurable. Thesis is impaired if Copilot attach rates, commercial RPO, or Azure AI growth decelerate materially for two consecutive quarters.
• Avoid using the survey as a catalyst to buy pure-play consumer AI exposure. Set an alert for independently reported paid-user retention, net adds, and inference-cost trends; without those data, satisfaction metrics do not establish durable unit economics.
• Watch META's next earnings call for incremental AI capex guidance versus disclosed revenue lift. A capex increase without evidence of ad ranking, conversion, or business-messaging monetization would support increasing the META underweight; quantified conversion lift would falsify it.