Americans are adopting AI chatbots faster than ever, but they’re also growing increasingly uneasy about how quickly the technology is moving. A new Pew Research poll, reported by The Verge, paints a picture of simultaneous momentum and mistrust: chatbot use is rising sharply, yet public confidence in AI’s direction and impact remains limited.
The headline numbers are striking. Forty-nine percent of Americans say they use AI chatbots at least occasionally. That figure matters not just because it’s close to half the population, but because it represents a major shift in a relatively short time. Pew’s findings show that chatbot use has climbed dramatically since 2024, when only 33% reported using them. In other words, within roughly a year, millions more people moved from curiosity or awareness into actual use—at least sometimes.
ChatGPT, specifically, appears to be a key driver of that adoption curve. Pew reports that ChatGPT usage has doubled since 2023, with 44% of respondents saying they’ve used it. For many Americans, ChatGPT isn’t simply another app; it’s become a kind of default entry point into the broader category of “AI.” That matters because early experiences often shape long-term attitudes. If the first encounter is helpful—drafting an email, summarizing a document, brainstorming ideas—people are more likely to keep using the tool. But if the experience is confusing, unreliable, or unsettling, skepticism can take root just as quickly.
And skepticism is exactly what Pew finds. Sixty-three percent of Americans say AI is advancing too quickly. That’s nearly two-thirds of the country expressing concern about the pace of progress. The poll doesn’t suggest that people are rejecting AI outright; rather, it suggests they’re questioning whether society is keeping up with the consequences. Adoption is happening in the foreground, while worry is building in the background.
This combination—more use, more concern—creates a tension that’s easy to miss if you only look at adoption rates. It’s tempting to interpret rising chatbot usage as proof that the public is comfortable with AI. But Pew’s results complicate that narrative. People can be curious and cautious at the same time. They can experiment without believing the technology is being deployed responsibly. They can benefit from AI tools while still feeling that the broader system is moving faster than governance, safety testing, and social adaptation.
One of the most revealing parts of the poll is how limited optimism appears to be. Only 16% of respondents say they believe AI will have a positive impact on society. That number is low enough to stand out immediately, especially when paired with the fact that nearly half of Americans already use chatbots. In practical terms, this suggests that many users don’t see AI as a net societal win—at least not yet, and perhaps not in the way they expected.
So what does it mean when people use something frequently but don’t expect it to improve society? It can mean several things, and Pew’s broader framing points toward a few likely interpretations.
First, people may be treating AI as a tool rather than a transformation. If someone uses a chatbot to help with a task—writing, studying, planning, troubleshooting—that doesn’t automatically translate into belief that AI will improve the world. A person can view AI as useful while still thinking it’s disruptive, risky, or unfairly distributed.
Second, the gap between personal utility and societal optimism may reflect uncertainty about downstream effects. Chatbots can feel straightforward in daily use, but their integration into workplaces, education systems, media ecosystems, and public services raises questions that aren’t resolved by a single good interaction. Even if a chatbot produces a helpful answer, users may still wonder: Who is accountable when it’s wrong? What happens when it’s used at scale? How do we prevent manipulation, bias, or fraud? How do we protect privacy? Those questions don’t always show up in the moment-to-moment experience of using a chatbot, but they can strongly influence beliefs about society.
Third, the “advancing too quickly” sentiment suggests that many Americans may feel the technology is outpacing the rules. When people think progress is too fast, they often imagine missing guardrails: insufficient regulation, inadequate transparency, weak oversight, and a lack of meaningful public input. That doesn’t necessarily stop adoption—if anything, it can coexist with it. People may try the technology because it’s already here, even if they don’t trust where it’s headed.
Pew’s findings also highlight a generational split that adds another layer to the story. The poll indicates that younger generations both report using AI more and hold more pessimistic views. This is important because it challenges a simplistic assumption that younger people are uniformly more optimistic about emerging technology. Instead, the data suggests that familiarity doesn’t automatically produce comfort. In some cases, the opposite may be true: the more people interact with AI, the more they notice its limitations, inconsistencies, and potential harms.
Why might younger Americans be both heavier users and more pessimistic? One possibility is that younger people are more exposed to AI’s social and cultural effects. They may encounter AI-generated content more frequently in everyday life—memes, images, text, and “synthetic” media—along with the confusion it can cause. They may also be more directly affected by AI’s implications for education and early career pathways, where automation and credentialing changes can feel immediate rather than theoretical.
Another possibility is that younger users are more likely to compare AI’s promises to its real-world performance. If you’ve used chatbots extensively, you may have seen them confidently produce incorrect information, fail to follow instructions, or struggle with context. Even when the tool is impressive, repeated exposure to errors can sharpen skepticism. Meanwhile, older users might rely on fewer interactions and therefore experience fewer moments that undermine trust.
There’s also a psychological angle worth considering: when a technology becomes mainstream quickly, it can trigger a “too fast” reaction. Younger people may have grown up with rapid tech cycles, but AI’s speed and scale may still feel unusually intense. If the technology seems to be reshaping communication, creativity, and work norms in real time, pessimism can emerge not because people are less capable of understanding AI, but because they understand it well enough to see the risks.
The poll’s results also invite a closer look at what “use” actually means. Pew reports that 49% use chatbots at least occasionally, which implies a spectrum—from casual experimentation to regular reliance. That matters because different levels of use can correlate with different attitudes. Someone who uses a chatbot once a month might be curious but not deeply invested. Someone who uses it daily might be more dependent, more aware of its quirks, and more sensitive to its failures. Without breaking down the attitudes by intensity of use, it’s hard to know exactly how the relationship plays out—but the overall pattern suggests that even among users, trust in AI’s societal impact is not strong.
This is where the story becomes more than a set of percentages. It becomes a snapshot of a society negotiating a new kind of technology adoption: one that is both accessible and consequential. Unlike earlier waves of tech—social media, smartphones, streaming—AI chatbots are not just platforms for content. They are systems that generate language, reasoning-like outputs, and decision-adjacent suggestions. That makes them feel more “agentic,” even when they are not truly autonomous. People may treat them as collaborators, assistants, or quasi-experts. And when a tool behaves like an expert, the stakes of being wrong rise.
That may help explain why only 16% believe AI will have a positive impact on society. If people perceive AI as influencing truth, judgment, and power—rather than merely assisting with convenience—then optimism requires more than personal usefulness. It requires confidence that the technology will be governed well, audited effectively, and aligned with public interests. Pew’s “advancing too quickly” result suggests that confidence is currently lacking.
At the same time, the poll doesn’t indicate that Americans are disengaging. If anything, the trend is upward. Chatbot use has increased substantially since 2024, and ChatGPT’s user base has expanded since 2023. That suggests that the market and the culture are pulling people toward AI regardless of concerns. In many households, AI is becoming part of the everyday toolkit—like search engines did, or like smartphones did. But unlike those earlier technologies, AI chatbots are more likely to produce persuasive text that can be mistaken for expertise. That difference could be central to the trust gap.
Consider the role of “confidence signals.” Chatbots often present answers in a confident tone, with fluent language that can make outputs feel authoritative. Even when users know the tool can be wrong, the presentation can still influence how people interpret the information. Over time, repeated exposure to plausible-sounding errors can create a specific kind of skepticism: not necessarily “AI is useless,” but “AI is unreliable enough that I can’t treat it as truth.” That skepticism can coexist with continued use, especially if people use chatbots as drafts, brainstorming partners, or starting points rather than final authorities.
The generational dimension also hints at how trust is being formed. Younger users may be learning these habits earlier—fact-checking, cross-referencing, and treating AI outputs as provisional. That could explain why they use AI more but remain pessimistic about society. They may see the tool’s limitations clearly, and they may also see how those limitations could scale into misinformation, unfairness, or job displacement.
There’s another angle: the poll’s results may reflect a broader cultural debate about responsibility. When a technology advances quickly, it forces institutions to respond quickly too—schools, employers, regulators, and platforms. If those institutions appear slow, inconsistent, or reactive, public trust can erode. People may feel that the technology is being adopted faster than the systems needed to manage it. That would align with the 63% who say AI is advancing too quickly.
In that sense, the poll captures a moment of transition. Americans are not simply deciding whether to adopt AI; they are deciding