New Pew Research Center data shows chatbot adoption remains deeply uneven by age and race, while privacy, accuracy and simple lack of interest keep millions of Americans from using the technology.
Nearly half of U.S. adults — 49% — say they never use AI chatbots such as ChatGPT, Gemini or Copilot, according to Pew Research Center’s latest analysis of Americans and artificial intelligence. The findings come from a Pew survey conducted February 17–23, 2026, and they cut against the impression that generative AI has already become a universal part of American life. Adoption is growing, but it remains highly uneven. Among adults 65 and older, 77% say they never use chatbots, compared with 58% of adults ages 50 to 64, 39% of those ages 30 to 49 and 34% of adults under 30. The divide also appears across racial and ethnic groups: 70% of Black adults, 51% of Hispanic adults and 54% of White adults say they do not use chatbots, compared with 29% of Asian adults surveyed. Pew’s findings suggest the next phase of AI adoption may be less about whether people can access the technology and more about whether they trust it, understand its value and see any reason to use it. Pew Research Center’s October 9 analysis examines who is staying away from chatbots and why.
The most striking finding is that nonuse is not primarily a skills problem. Among Americans who never use chatbots, 60% say lack of interest is a major reason, making it the largest barrier Pew measured. Privacy follows closely, with 54% saying concern about how their personal information will be used is a major factor, while 45% cite doubts about whether chatbots provide accurate information. Only 29% say not knowing how to use the technology is a major reason. That changes the challenge facing AI companies. If the primary barrier were technical literacy, companies could respond with better tutorials, simpler interfaces or more training. But persuading someone who does not trust the product — or does not think the product is useful — requires something more fundamental than improved onboarding.
That skepticism also appears unlikely to disappear quickly. Pew found that 67% of current chatbot non-users say they are unlikely to use one during the next 12 months, including 40% who say they are not at all likely to start. Only 5% say they are very or extremely likely to begin using a chatbot, while another 16% say they are somewhat likely. About 12% remain unsure. This is important because much of the technology industry continues to talk about artificial intelligence as though mass adoption is inevitable. The Pew data suggests otherwise. A significant share of the country is not waiting for a better model or a lower price; many people have already decided the technology is either unnecessary, untrustworthy or not useful enough to justify changing their behavior.
The Age Divide Could Become a Workplace Divide
The age gap matters because AI is moving beyond optional consumer experimentation and into the infrastructure of work. Microsoft, Google and other technology companies are integrating AI assistants into productivity software, search, coding tools and enterprise platforms. Employers are increasingly asking workers to use AI for writing, analysis, customer service, research and administrative tasks. Younger adults who already use chatbots regularly may therefore enter workplaces with a growing familiarity advantage, while older employees who have avoided the technology could encounter it for the first time through an employer mandate rather than personal choice.
That does not mean workers over 50 are incapable of adapting, nor does it mean chatbot use automatically translates into greater productivity. But the gap creates another version of the digital divide: not simply access to technology, but familiarity with the technology before it becomes a job expectation. More than three-quarters of Americans 65 and older and a majority of adults between 50 and 64 currently do not use chatbots, according to Pew. If AI becomes embedded in more occupations, employers may eventually have to treat AI literacy the way organizations once approached spreadsheet, email and internet literacy — as a skill that needs to be taught rather than assumed.
The Racial Adoption Gap Deserves More Attention
The racial differences are also difficult to ignore. Pew found 70% of Black adults say they never use chatbots, the highest nonuse rate among the racial and ethnic groups included in the survey. That compares with 54% of White adults, 51% of Hispanic adults and 29% of Asian adults. Pew’s survey does not establish a single reason for those gaps, and the data should not be interpreted as evidence that race itself determines whether someone adopts AI. Age, occupation, education, exposure to technology, trust, income and perceived usefulness can all shape adoption patterns.
But the consequences could become significant if generative AI increasingly influences education, hiring, entrepreneurship and professional productivity. A technology can technically be available to everyone while still producing unequal benefits if some groups are less likely to use it. We have seen versions of that pattern before with broadband access, personal computers and other technologies that eventually became intertwined with economic mobility. The risk is not that everyone must use ChatGPT or another chatbot. It is that the economic value of AI could become concentrated among the people and workplaces that learn how to use it first.
That possibility is particularly important because AI companies increasingly market their products as tools for opportunity. Chatbots can help users draft résumés, prepare for interviews, learn technical skills, build businesses, research markets or navigate unfamiliar professional environments. If those functions become genuinely useful, then unequal adoption becomes more than a consumer preference. It becomes a question about who is positioned to capture the benefits of another major technological shift.
AI Companies Have a Trust Problem
The privacy and accuracy findings may be the hardest for the industry to dismiss. Among chatbot non-users, 79% say concerns about personal-data use are at least a major or minor reason they stay away, while 76% say doubts about accuracy are at least part of the reason. Those concerns reflect two of the most persistent problems surrounding generative AI: what happens to the information users provide and whether the answers they receive can be trusted.
Chatbots have improved considerably, but hallucinations have not disappeared. The systems can still produce incorrect information confidently, creating a different user experience from traditional search, where people are generally shown sources they can independently evaluate. AI developers have also faced recurring questions about training data, stored conversations, personalization and how user information may be processed. For people already skeptical of large technology platforms, asking them to place private questions, professional information or personal decisions inside another digital system can be a significant hurdle.
That trust problem will become more important as companies try to move chatbots into higher-stakes areas. Writing an email or summarizing a document carries one level of risk. Using AI to interpret medical information, evaluate financial choices, assist with schoolwork or make employment decisions carries another. The more ambitious the use case becomes, the more accuracy and privacy stop being technical side issues and become central to whether people will use the product at all.
The Industry May Be Overestimating Inevitability
The most important takeaway from Pew’s research may be that AI adoption is not following one smooth curve toward universal use. There are still more Americans who say they never use chatbots than Americans who say they do, and most of the people on the sidelines do not currently expect to cross over. That does not mean adoption has stalled permanently. Smartphones, social media and online banking all encountered periods when large segments of the public resisted them before eventually adopting them. But AI companies should not assume the same trajectory is guaranteed.
The technology industry has largely competed on capability: whose model reasons better, writes faster, generates stronger images or integrates with more software. Pew’s data suggests the next competitive problem may be more human. Companies have to demonstrate why these tools deserve a place in people’s lives. They have to convince skeptical users that their information is safe, that answers can be trusted and that chatbot use solves a meaningful problem rather than creating another digital task to manage.
That makes the divide Pew identified more consequential than a simple measure of technology adoption. Artificial intelligence is simultaneously moving deeper into workplaces, schools, government services and consumer platforms while roughly half of Americans remain outside the chatbot ecosystem. If that divide persists, the country could reach a point where AI is commonplace inside institutions without being commonplace among the people those institutions serve. The next chapter of the AI story therefore will not be determined only by what the models can do. It will also depend on who decides to use them, who continues to opt out and whether that choice eventually carries an economic cost.