In 2024, 33% of American adults used chatbots. Two years later, the same institution asked the same question, and the number had jumped to 49%. Alongside those figures came another: 63% of respondents said AI is advancing too quickly. That number is higher than the adoption rate itself.

Pew Research Center addressed this gap directly in its June 2026 report, "Americans and AI." The phrase "growth in usage does not equal trust in the technology" appeared in the report's body text for the first time. The researchers chose to weight interpretation over raw numbers.

The Devices Spread. The Trust Didn't.

The report organized the AI landscape around four axes: the mainstreaming of chatbots and AI-powered search, the penetration of smart home devices, growing skepticism about data security and societal impact, and widening perception gaps across demographic groups.

On the adoption side, the spread is unmistakable. Beyond the 16-percentage-point jump in chatbot usage over two years, 60% of adults now read search results that surface AI-generated summaries at the top. AI-powered devices like smartwatches (37%) and smart speakers (35%) have become fixtures in American households.

The psychological landscape moved in a different direction. Seventy-one percent said AI will make personal data less secure. Negative views on AI's societal impact (40%) outpaced positive ones (16%) by more than two to one. The share of Americans who don't trust the government to regulate AI effectively rose from 62% in 2024 to 67%. The share who don't trust AI companies to handle the technology responsibly held steady at 60%.

There's a specific discomfort in seeing these numbers side by side. People are actively using the tools while trust in the people building them isn't rising — it's falling.

Cross-referencing the data produces an even sharper picture. Sixty-six percent of 18-to-29-year-olds use chatbots — the highest rate of any age group. But this same cohort also reported the highest rate of concern about AI's societal impact, at 48%. By contrast, Asian Americans led on both usage (70%) and positive perceptions of AI's personal benefits (41%). The same tool reads very differently depending on who's holding it.

Skeptical Use May Not Be a Contradiction

How to read this paradox depends on your vantage point.

One view is the exposure effect. The more you use AI, the more you witness its errors and limits firsthand. When a chatbot confidently produces something false, or veers in an unexpected direction in a sensitive context, distrust hardens into something far more concrete than vague unease. The anxiety of someone who has never tried the tool is different in kind from the skepticism of someone who uses it every day.

There's also an information-access interpretation. Heavy AI users tend to encounter more AI-related news and debate. When stories about algorithmic bias, unauthorized data collection, and regulatory gaps flow through your feed, they're harder to look away from than they would be for a casual user. From this angle, the finding that distrust is highest among heavy users isn't a product of ignorance — it's a product of knowing more.

A critical reading finds something unresolved in the survey itself. "Don't trust" responses don't cleanly distinguish between distrust of AI technology in general, of specific companies' products, or of the government's capacity to regulate it. When "distrust" is defined too broadly, the explanatory power of the number weakens. And measuring usage and distrust simultaneously doesn't establish causation. It's equally plausible that people with naturally critical dispositions are the ones most inclined to try the tools themselves.

Even so, one pattern the report captures is hard to dismiss. The behavior Gen Z displays — using AI most actively while questioning it most sharply — looks, from the standpoint of how to handle a tool, like a healthy posture. Convenience doesn't vouch for correctness. People who remain conscious of that gap may be the ones who ultimately use tools most effectively.

When I first encountered this data, I read Gen Z's high distrust as a negative signal. But it can be read differently. If the heaviest users are also the most skeptical, they may be the group that has already learned from experience exactly where and how AI gets things wrong — and is moving toward selective, informed use rather than uncritical adoption.

What Practitioners Can Actually Do With These Numbers

This survey was conducted with American adults. The numbers don't translate directly to other workplaces. Platforms, work culture, and regulatory context differ by country. But the structural gap — adoption outpacing trust — is likely shared across markets.

AI tool use is growing steadily in workplaces everywhere. Chatbots have moved into meeting summaries, proposal drafts, customer-facing copy, and market research. Alongside that, questions like "Can I use this output as-is?" and "Is it okay to put my name on something AI wrote?" come up regularly on the ground.

In this context, what practitioners should examine isn't how often they use AI — it's what criteria guide their judgment. Frequent use doesn't mean accepting everything AI produces. Consciously deciding which tasks to delegate, which outputs to review, and where human judgment must step in is what real tool competency looks like.

Research on the future of work consistently points in the same direction. As the range of tasks AI handles expands, the human role shifts toward a layer of decisions: when to intervene, when to hand off. Knowing when not to use a technology will become rarer, and more valuable, than knowing how to use it.

The finding that 71% believe AI makes personal data less secure carries its own practical implication: you need to anticipate how customers or readers will respond psychologically to AI-generated content or recommendations. That wariness will run higher in sensitive domains — healthcare, finance, law. For solo practitioners and small teams, how you explain your use of AI to clients matters as much as how you use it. Whether "we use AI" reads as efficiency or as distance depends entirely on context. Reading that context is one of the most underaddressed elements of AI adoption right now.

An organization that has deployed AI and concludes "our team is using it well" may be missing the fact that those same team members are using it while actively doubting it. Pew's diagnosis — that usage rates don't reflect the state of trust — operates the same way inside organizations.

More users doesn't mean the tool has properly taken root. How people are using it, where they stop, how many double-check the output — those are the questions that outlast the adoption numbers.