Does More AI Use Mean More Trust in AI?

More people using a technology does not necessarily mean that they trust it. Use and trust are related, but they answer different questions: use records whether someone chooses or needs to interact with a system, while trust concerns whether the person believes the system is reliable, appropriate, and safe enough to depend on. A person may use an AI chatbot to complete a task while checking its answer elsewhere. A driver may accept a warning from a vehicle system without believing the system can handle every hazard. Adoption can create opportunities for confidence to grow, but it is not itself proof of confidence.

Pew Research Center’s 2025 survey makes clear that AI chatbots have become part of many U.S. teenagers’ technology use. The survey found that roughly two-thirds of teens reported using chatbots, and about three in ten reported daily use. That is meaningful evidence of adoption: many teens have tried these tools, and a substantial group encounters them frequently. But those figures measure reported behavior, not trust. They do not tell us whether respondents believe chatbot answers, feel comfortable sharing information, or would rely on a chatbot for an important decision. The survey therefore supports the claim that use is widespread, not the separate claim that confidence is widespread. (Pew Research Center)

The difference matters because people can use a system for reasons that have little to do with trusting it. A chatbot may be convenient, entertaining, or simply available when a person wants a quick response. Someone can also use it experimentally, to see what it can do, while remaining skeptical about the result. Daily use makes familiarity more plausible, but familiarity is not the same as confidence: repeated interaction can teach a user when a tool is helpful and when it fails. Pew’s figures establish how often teens say they use chatbots; without a direct measure of confidence or reliance, they cannot establish why teens use them or how much they trust them.

NHTSA’s descriptions of driver-assistance technology offer a concrete comparison between interacting with automated systems and relying on them. A forward-collision warning uses sensors to detect slower or stopped vehicles and alerts the driver when a crash may be imminent, giving the driver an opportunity to brake or steer. A lane-departure warning similarly alerts a driver to an unintentional drift, but NHTSA emphasizes that this warning system does not itself take action to avoid a crash. These examples show that a person can receive and use a system’s information while retaining responsibility for judging what to do. The presence of technology in a routine task does not mean that a driver has handed it complete authority.

Other NHTSA-described systems intervene more directly, which makes the distinction between use and trust even more important. Automatic emergency braking can apply brakes to avoid or reduce the severity of an impending forward crash. In one described situation, dynamic brake support supplements a driver’s braking if it is not hard enough; in another, crash imminent braking applies the brakes if the driver does not brake. NHTSA says research on these systems and relevant performance measures found that some systems available in the marketplace can avoid or reduce the severity of rear-end crashes in certain situations. This evidence supports a qualified case for relying on specific safety functions under particular conditions. It does not establish that every user trusts every automated feature, or that any system works in every situation.

A serious counterargument is that frequent use can be a sign of trust. Someone who uses a chatbot daily or permits a car to assist with braking may be demonstrating at least some practical confidence; otherwise, the person might avoid the system. Repeated successful interactions could also strengthen confidence over time. This argument is reasonable, especially when use involves a deliberate decision to depend on a tool rather than simply encountering it. Yet it shows only that use may provide clues about trust, not that it measures trust directly. A person might use a tool while checking its output, limiting it to low-stakes tasks, or relying on it because it is convenient.

The limits of the survey also matter. Pew’s study was an online survey of 1,458 U.S. teens ages 13 to 17, conducted from September 25 to October 9, 2025. Its findings concern teens’ reported use during that period; they should not automatically be generalized to adults, people in other countries, or later years. The survey was weighted to represent U.S. teens ages 13 to 17 living with their parents, and its method provides a basis for describing that population. Still, a use question cannot answer an unasked question about trust. To investigate confidence, researchers would need to ask directly what people trust, for which tasks, and under what conditions.

The best conclusion is therefore neither that rising use proves trust nor that use tells us nothing about it. Adoption shows that a technology has entered people’s routines and may reflect convenience, access, curiosity, practical need, or some degree of confidence. Pew’s chatbot survey documents substantial use among U.S. teens, while NHTSA’s account shows how vehicle systems can support drivers in defined safety situations—and why the functions and limits of individual systems matter. These examples make use relevant evidence, but incomplete evidence, about trust. To claim that people trust AI more, we need measures of their confidence and reliance, not just counts of how often they use it.

How the reasoning works

The essay defines adoption and trust separately, then uses Pew’s survey to establish chatbot use without treating it as a measure of confidence. NHTSA’s descriptions supply a distinct U.S. example: driver-assistance systems vary in how they alert or intervene, and their benefits are described for specific situations. The counterargument recognizes that deliberate, repeated use can suggest practical confidence, then explains why that inference remains incomplete. The final paragraphs qualify the survey’s population and period and identify what additional evidence about trust would be needed.