The Unease Paradox

Roughly two-thirds of Americans say AI is advancing too quickly — even as adoption of chatbots and AI tools has roughly doubled over two years. That is the central tension in Pew Research Center's "Americans and AI 2026" report, published June 17, 2026.

This concern spans demographics. Pew's breakdown by age found majorities across all adult age brackets share the worry, ranging from 61% to higher depending on generation. The gender split is more pronounced: 68% of women versus 58% of men say the pace is too fast — a ten-point gap that has appeared consistently in Pew's AI tracking. The gap likely reflects differing exposure to AI systems in content moderation, hiring tools, and surveillance applications, which have drawn the most scrutiny.

Alongside speed anxiety, majorities also told Pew they believe AI poses a risk to their personal information. This pairing — pace concern and data vulnerability — mirrors how public sentiment formed around earlier platform shifts. Search engine personalization, social media data collection, and smartphone location tracking each generated similar dual anxieties before market or regulatory pressure forced at least partial changes.

Usage Is Climbing Steeply

Yet usage numbers tell a different story. A 2025 Pew survey found 34% of US adults had used ChatGPT — roughly double the 2023 figure. Among adults under 30, the figure reaches 58%. By mid-2026, Pew data shows chatbot usage among Americans under 50 at around 63%, suggesting the adoption curve has accelerated further.

Teens are a notable cohort. By early 2025, 26% of US teens aged 13 to 17 reported using ChatGPT for schoolwork — up from 13% in 2023, a doubling in roughly one school year. For context, that rate of education-relevant adoption is faster than what was seen in the early years of tablet deployment in classrooms or consumer search engines finding their way into homework. K-12 institutions are still sorting out policies, with approaches ranging from full bans to formal integration varying by school district.

Pew also recorded broader growth in AI summary features and smart speaker usage, consistent with AI expanding beyond explicit chatbot sessions into ambient and inline interfaces — search results, voice assistants, email drafts — where users may not consciously register the interaction as AI at all.

The Gap Between Concern and Behavior

The divergence between stated anxiety and actual usage is not a contradiction so much as a recurring feature of how people adopt new platforms. People routinely use technologies they distrust at an institutional level while finding individual utility compelling enough to override that distrust. Social media is the most documented recent example; ride-hailing services, with their privacy and labor concerns, is another.

What distinguishes the current AI moment is how compressed the cycle has become. The time from ChatGPT's release in late 2022 to a third of US adults having used it was roughly two and a half years. By comparison, smartphones took closer to six years after the iPhone's 2007 launch to reach equivalent household penetration, and broadband internet took even longer. The adoption curve for AI-native applications is plainly faster.

That speed is exactly what the Pew data captures in how Americans frame their concern. When nearly two-thirds say AI is advancing too quickly, they are not necessarily calling for a halt — the usage numbers make that implausible. They are more likely expressing that the social, regulatory, and institutional structures have not kept pace. Schools are still drafting chatbot policies. Congress has not passed comprehensive AI legislation. Legal frameworks for AI-generated outputs remain unsettled.

For anyone building AI products, deploying models in enterprise settings, or advising on AI governance, the practical message of this survey is direct: public appetite for AI tools remains high enough to sustain adoption, but institutional trust in how AI is governed has not matched that appetite. That gap — between how much people value individual tools and how much confidence they have in the broader systems managing AI — is where policy and technical work will concentrate over the next several years.