About half of Americans now use AI chatbots — and most are deeply skeptical of where AI is headed.
Chatbot use has climbed from a third of U.S. adults in 2024 to about half today, AI is moving into people's homes through smart speakers and watches, and AI summaries now sit atop most people's search results. Yet across nearly every measure of sentiment, Americans expect AI to do more harm than good — to society, to their privacy, and to a future they think is arriving too fast. A reader's map of the nine threads in Pew's February 2026 survey of 5,119 adults — plus a tenth on what has happened since, and an eleventh placing Americans among 37 countries.
Nine threads · one survey — then what came next, and how the U.S. compares
01Half of U.S. adults now use chatbots
About half of Americans report ever using an AI chatbot like ChatGPT, Gemini or Copilot — up substantially from roughly a third in the summer of 2024. Awareness of AI is now nearly universal, even as actual daily use remains concentrated.
Roughly one in four adults use chatbots on a daily basis — including 12% who use them several times a day and 4% who use them almost constantly. Another quarter report using these tools several times a week or less, and about half of adults don't use chatbots at all.
Awareness has raced ahead of use. 96% of adults now say they know at least a little about AI, and the share who have heard "a lot" has grown to about half — up from just 26% in 2022. For chatbots specifically, roughly nine in ten have heard at least a little (including 44% who have heard a lot), compared with about seven in ten in August 2024.
Confidence, though, lags behind awareness: only about 1 in 5 adults say they are highly (extremely or very) confident in their ability to use AI chatbots.
How often U.S. adults use AI chatbots
The thread
Use roughly doubled in under two years, but the population splits cleanly in half: a daily-and-weekly user base of about 50%, and a non-using half. Awareness is no longer the barrier — familiarity is near-universal while confidence and habitual use are not.
02ChatGPT dominates a crowded field
Since ChatGPT's 2022 debut, several rivals have entered the market — but ChatGPT still towers over them. A little under half of U.S. adults (44%) now report using it, up from 34% a year earlier and more than double its 2023 share.
% of U.S. adults who ever use each chatbot
The second-most-used platform is Gemini, used by about a quarter of adults (24%), followed by Copilot (17%) and Meta AI (14%). About one in ten or fewer say they use Grok (8%), Claude (6%) or Character.ai (3%).
ChatGPT's reach is still growing: its 44% share is up from 34% the prior year and more than double its level when Pew first asked in 2023.
Use varies sharply by age. Adults under 50 are about twice as likely as those 50 and older to use ChatGPT — 57% vs. 28%. (The full age-by-brand breakdown appears in Section 08.)
The thread
This is a one-brand market with a long tail. ChatGPT alone is used by nearly as many adults as every other chatbot combined, and it's still the fastest-growing. Brand recognition and adoption are not yet evenly distributed across the field.
03Search and work lead; support is the long tail
Americans reach for chatbots across many parts of daily life, but two uses dominate: searching for information and getting work done. The more personal uses — emotional support and companionship — remain a small minority, though they fuel an outsized share of the public debate.
42% of adults use chatbots to search for information, and 38% of employed adults use them for tasks at work. From there it steps down: 25% for fun or entertainment, 24% to create or edit images or videos, and 20% each for medical advice and for diet and fitness information.
Further down: 13% use chatbots to get news. And in the categories that draw the most scrutiny, 10% report using chatbots for emotional support or advice, while just 4% do so for companionship.
These patterns vary widely by age, gender and race/ethnicity — for example, 1 in 5 adults under 30 use chatbots for emotional support, roughly double the rate among 30-to-49-year-olds. See Section 08.
% of U.S. adults who ever use chatbots for…
The thread
Chatbots are mostly an information-and-productivity tool: people use them to look things up and to work. The headline-grabbing uses — therapy-like support, companionship — are real but rare in aggregate, and concentrated among the youngest adults.
04Help more than hurt — but only modestly
Among those who use chatbots, the perceived effects on everyday life tilt positive for productivity, knowledge and creativity. But the dominant answer for happiness and relationships is that chatbots make no difference at all.
Among chatbot users: helps vs. hurts each area
Americans who use chatbots are more likely to say the tools help rather than hurt their productivity and how informed they are. About three in ten say chatbots help their productivity, and a similar share say chatbots help keep them informed. Just 5% each say chatbots hurt them in those areas.
The picture for creativity is also net-positive but closer: about one in five say chatbots help their creativity, versus 11% who say they hurt it.
For happiness and relationships, the most common answer is that chatbots neither help nor hurt. Few users report any meaningful effect in either direction on their personal connections or wellbeing.
The thread
Where chatbots touch tasks — getting work done, staying informed, generating ideas — users see net benefit. Where they touch feelings and relationships, users mostly see no effect. The tool is perceived as useful, not transformative, for personal life.
05AI in the home — and at the top of search
Chatbots are only one way AI shows up in daily life. From Alexa speakers to Apple Watches, AI is embedded in everyday devices — and AI-generated summaries now greet a majority of Americans at the top of their search results.
About four in ten adults have a smartwatch, while a majority say they don't. Inside the home, roughly a third have a smart speaker — the most common AI device — followed by smart doorbells (about one in five) and robot vacuums or smart thermostats (about one in ten each).
AI is also reshaping how people consume information. Six in ten adults say they read the AI summaries that now appear at the top of search engine results. About three in ten say they do not, and another 10% aren't sure whether they have.
Pew notes these conveniences come with concerns the survey probes elsewhere — surveillance and child safety among them.
Do you read AI summaries atop search results?
The thread
AI adoption isn't only deliberate. People choose a chatbot, but a smart speaker, a watch, or an AI search summary brings AI into their lives more passively. A majority now read AI summaries — arguably the most widespread AI touchpoint in the survey.
06Skeptical of the future, worried about privacy
As more Americans fold AI into their lives, their outlook stays wary. More predict AI will harm rather than help society and themselves, large majorities think it's advancing too quickly, and most expect it to make their personal information less secure.
Impact of AI over the next 20 years
Societal impact tilts clearly negative. Four in ten adults say AI will have a negative impact on society over the next 20 years; far fewer expect a positive one. Views on personal impact tilt negative too, but less sharply — 31% expect a negative effect on their own lives versus 23% who expect a positive one. About three in ten think the effects will be equally positive and negative (31% for society, 27% for themselves).
Speed. About two-thirds of Americans say AI is advancing too quickly. Only 2% say too slowly, about one in five say it's moving at about the right pace, and a similar share aren't sure.
Data security. Roughly seven in ten predict AI will make their personal information less secure. Just 3% think it will make their information more secure; about one in ten say it won't make much difference, and 16% are unsure.
The thread
Sentiment is consistently defensive: negative on net impact, alarmed about pace, and near-unanimous that privacy will suffer. Notably, people are more pessimistic about society than about themselves — a pattern where individuals feel they can manage AI even as they fear its broader effects.
07Low confidence in both government and companies
The skepticism extends to the institutions shaping AI. Majorities aren't confident that government is regulating AI effectively or that companies are building it responsibly — and the partisan map of that distrust has flipped.
67% of Americans have little to no confidence in the U.S. government to regulate AI effectively — slightly up from 62% in 2024. And about six in ten adults aren't confident that U.S. companies will develop and use AI responsibly.
The partisan reversal. Distrust is common in both parties, but Democrats are now more skeptical than Republicans of government's ability to regulate AI — 74% vs. 61% (each group including leaners). That's a shift: Republican lack of confidence fell from 70% in 2024 to 61% today, while Democratic lack of confidence rose 20 points over the same span.
On companies, a larger share of Democrats than Republicans lack confidence in responsible AI development (65% vs. 53%) — a gap that didn't meaningfully exist two years ago.
% with little/no confidence to regulate AI
The thread
No major institution earns the public's confidence on AI. The striking movement is partisan: as the national conversation and administration shifted, Democrats and Republicans essentially swapped positions on trusting government to regulate AI, with Democrats now the more skeptical group.
08Age, gender, and race/ethnicity
Use and opinion are far from uniform. Younger adults use AI the most yet trust it the least; men use it more heavily than women, who are more skeptical; and Asian adults stand out for both adoption and optimism.
By age — heavy use, high skepticism
Adults under 50 are far more likely to use chatbots: about 63% of under-50s use them, versus roughly four in ten of those 50–64 and fewer of those 65+. A third of under-50 adults use chatbots daily. Yet younger adults are the most wary: roughly half of adults under 30 say AI will negatively affect society (just 14% positive), and about four in ten say it will be negative for them personally — more pessimistic than older groups. Older adults are instead the most unsure. Across every age group, 61–65% say AI is advancing too quickly.
| Chatbot | All adults | 18–29 | 30–49 | 50–64 | 65+ |
|---|---|---|---|---|---|
| ChatGPT | 44 | 61 | 55 | 37 | 19 |
| Gemini | 24 | 30 | 34 | 20 | 11 |
| Copilot | 17 | 17 | 22 | 19 | 10 |
| Meta AI | 14 | 14 | 17 | 13 | 10 |
| Grok | 8 | 9 | 10 | 7 | 3 |
| Claude | 6 | 10 | 9 | 4 | 1 |
| Character.ai | 3 | 7 | 3 | 1 | 1 |
Not all differences shown are statistically significant. Other age patterns: 1 in 5 adults under 30 ever use chatbots for emotional support (vs. ~1 in 10 of those 30–49); ~30% of under-30s feel highly confident using chatbots, versus 6% of those 65+.
By gender — men use more, women more skeptical
The overall gender gap in chatbot use has closed — men and women are now equally likely to say they use chatbots, a shift from 2024 when men led by 11 points. But men use them more regularly (27% daily vs. 20%) and for more purposes, and they're more likely to use specific brands like Copilot (22% vs. 13%), Gemini, Grok and Claude. ChatGPT is used by identical shares. Men are also more likely to say chatbots help their productivity (35% vs. 25%), to have heard "a lot" about AI (55% vs. 41%) and chatbots (48% vs. 38%), and to feel confident using them (22% vs. 15%). Women are slightly more likely to own a smartwatch (40% vs. 34%).
| Measure | Men | Women |
|---|---|---|
| Use chatbots daily | 27 | 20 |
| Ever use Copilot | 22 | 13 |
| Say chatbots help their productivity | 35 | 25 |
| Heard "a lot" about AI | 55 | 41 |
| Read AI search summaries | 63 | 57 |
| Say AI is advancing too quickly | 58 | 68 |
| Expect AI to be negative for them personally | 27 | 33 |
Women are about twice as likely to say AI will be negative rather than positive for them personally (33% vs. 17%); men are roughly evenly split. 43% of women expect AI to harm society. Women are also more likely than men to say AI is moving too fast (68% vs. 58%).
By race & ethnicity — Asian adults stand out
Asian adults are distinctive for both use and optimism. Seven in ten Asian adults use chatbots (vs. about half or fewer of Hispanic, Black and White adults), roughly half use them daily, and about seven in ten have heard "a lot" about chatbots. Asian adults are also the only group whose views on AI's personal impact tilt positive — about twice as likely to expect a positive rather than negative personal effect (41% vs. 20%) — and the most positive about societal impact. Still, majorities of every group (from more than half of Black adults to about six in ten or more of others) agree AI is advancing too quickly.
| Use | White | Hispanic | Black | Asian* |
|---|---|---|---|---|
| Search for information | 39 | 41 | 42 | 66 |
| Tasks at work (employed) | 36 | 37 | 36 | 60 |
| Fun or entertainment | 22 | 27 | 28 | 35 |
| Create or edit images/videos | 22 | 24 | 23 | 38 |
| Get medical advice | 18 | 23 | 21 | 39 |
| Diet and fitness information | 16 | 25 | 21 | 31 |
| Get news | 10 | 17 | 18 | 29 |
| Emotional support or advice | 7 | 13 | 15 | 17 |
| Companionship | 2 | 7 | 5 | 9 |
*Estimates for Asian adults are representative of English speakers only. White, Black and Asian adults are single-race and non-Hispanic; Hispanics are of any race. Not all differences shown are statistically significant.
The thread
The demographics complicate any simple story. The heaviest users are not the biggest believers. Young adults use AI most but fear it most; men use it more than women, who are warier; and Asian adults are unique in pairing the highest adoption with the most optimism.
09How Pew conducted the survey
The findings come from Wave 187 of Pew Research Center's American Trends Panel (ATP), a nationally representative panel of randomly selected U.S. adults, fielded Feb. 17–23, 2026.
The panel. Since 2018 the ATP has recruited through address-based sampling (ABS): study materials are mailed to a stratified random sample of households drawn from the U.S. Postal Service's Computerized Delivery Sequence File (estimated to cover 90–98% of the population), with the next-birthday adult in each household selected. A national sample has been recruited roughly once a year since 2014.
Asian oversample. This survey deliberately oversampled non-Hispanic Asian adults — selected with certainty — to produce more precise estimates for this smaller subgroup, then weighted them back to their true population share. Estimates for Asian adults represent English speakers only.
Weighting. The data is weighted in stages: a base weight for each panelist's recruitment probability, calibrated to population benchmarks to correct for nonresponse and attrition, then re-calibrated among completers and trimmed at the 1st and 99th percentiles to limit precision loss. Tests of statistical significance account for the weighting.
Data quality. Researchers screened for "satisficing" — blank answers or always picking the first or last option — and removed seven respondents before weighting and analysis. The break-off rate among those who started the survey was under 1%.
Beyond sampling error, question wording and the practical difficulties of conducting surveys can introduce additional error or bias. Income tiers in the study are adjusted for household size and regional cost of living using BEA Regional Price Parities.
The thread
This is a large, probability-based national sample with a tight margin of error (±1.6 points) and a deliberate Asian oversample — which is why the race/ethnicity breakdowns in Section 08 can be reported with confidence, with the caveat that Asian estimates cover English speakers only.
10Six months on, the mood hardened
Not drawn from Wave 187. Sources are named inline throughout.
Everything above was measured in February. By August the picture had moved twice: Pew fielded a later survey showing concern still climbing and job fears spreading, and the argument shifted from whether the public is uneasy to what that unease is starting to cost.
A separate, later Pew wave — 3,488 U.S. adults, fielded June 22–28 and published Aug. 18 — found 52% of Americans now say they are more concerned than excited about the increased use of AI in daily life, up from 37% in 2021. Only about one in ten say they are more excited than concerned.
The sharper movement is on jobs. 71% now think AI will lead to fewer jobs in the U.S. over the next two decades, up from 64% in 2024. Just 5% think it will create more — unchanged. Every age group moved, but the steepest jump is among 18-to-29-year-olds, up 12 points to 73%.
That pattern echoes Section 08: the age groups using these tools most heavily are also the ones whose expectations are darkening fastest. Concern is highest among adults 65 and older (59%) and lowest among those 50 to 64 (47%), with 18-to-29-year-olds at 55%.
"AI will mean fewer jobs in 20 years"
Where the sentiment is landing
TechCrunch's argument is that souring public opinion has stopped being a communications problem and started being a business one. Four strands of reporting, from four different outlets, point the same direction:
Alongside the balance-sheet effects runs a cultural one: young people adopting dumbphones, point-and-shoot cameras and tape decks; AI-free classic iPods fetching top dollar on eBay; quilting, jigsaw puzzles and Mahjong turning up everywhere; in-person run clubs winning out over dating apps. Opting out is becoming a consumer category in its own right.
The industry starts to answer
Two of the more candid responses came from chief executives who accept the premise rather than dispute it — one framing the gap as a product failure, the other as a failure to deliver on what was promised.
Amodei framed the broader problem as a "crisis of trust" — arguing that people don't trust companies, governments or the tech industry generally, and "suspect that we are cooking up some new way to screw them over." His proposed remedy is to actually deliver on the promises, curing cancer among them. Both quotes are reported by TechCrunch; the underlying podcast appearance and X post are the primary sources.
The thread
Sections 01–09 establish a gap between adoption and trust. This section is what that gap costs once it persists: it stops being a survey finding and starts being a constraint — on where data centers get built, on which party owns the issue, on what a founder's name is worth. The most striking thing is not that the public is skeptical but that the industry has begun to concede the point, and to locate the problem in what has been shipped rather than in how it has been explained.
11Among 37 countries, Americans sit near the gloomy end
Pew Spring 2026 Global Attitudes Survey — 42,151 adults in 36 countries, fielded Feb. 8–May 13, 2026, published Sept. 17 — read against the two U.S. waves above. Not drawn from any new U.S. sample.
Pew put the same questions on jobs, inequality, concern and awareness to adults in 36 other countries. It did not field a new U.S. survey: the American figures in the global release are the February wave (awareness) and the June wave (everything else) already on this page. So this section revises no U.S. number. It supplies the one thing a single-country survey can't — a sense of whether American unease is ordinary or unusual. It is unusual.
Jobs. In 34 of the 37 countries, more people expect AI to mean fewer jobs over the next 20 years than more. Only Australia and South Korea (76% each) are more pessimistic than the U.S. (71%); the 37-country median is 46%. Very few anywhere expect AI to add jobs — the U.S. figure is 5%, the median 9% — and the only places where a fifth or more say so are middle-income: Ghana (29%), Nigeria and Bangladesh (27%), Sri Lanka (23%), Pakistan (22%), India (21%).
The chart sorts itself by wealth. The eight most pessimistic countries are all high-income; the seven least pessimistic are all middle-income. Pew attributes much of the gap to uncertainty rather than optimism: in middle-income countries a median 34% say they are not sure how AI will affect jobs (54% in the Philippines, 49% in Thailand), against 22% in high-income ones and 14% in the U.S.
Inequality. Asked whether AI will widen or narrow the gap between rich and poor, the U.S. again ranks third (46% widen), behind South Korea (54%) and Australia (52%); the median is 28%. Nowhere do more than a quarter think AI will narrow the gap, and in 29 countries the figure is in single digits. Two U.S. splits here are new to this page: Americans under 35 are far more likely than those 50 and older to expect a wider gap (56% vs. 38%), and liberals are 31 points more likely than conservatives to say so — among the largest ideological gaps Pew found in any country.
Concern. Across the 37, a median 37% are more concerned than excited about AI in daily life and 41% feel both equally — the most common answer in about half the countries. The U.S. (52%) is second only to Australia (53%), level with Brazil. Israel is the only country where the primarily excited (29%) outnumber the primarily concerned (19%).
"AI will lead to fewer jobs" — % by country
The U.S. against the 37-country median
High-income vs. middle-income medians
A paradox in the pattern
Put the charts side by side and two findings seem to pull against each other. Between countries, the richer and better-informed publics are the most worried: high-income countries have nearly twice the share who have heard a lot about AI (50% vs. 27%) and markedly more who expect job losses and a wider wealth gap. Within countries, the relationship often runs the other way — people who have heard a lot about AI tend to be less alarmed than those who have heard little.
Germany: concern by how much people have heard
Both can be true at once. What drives the cross-country gradient is largely the size of the "not sure" group, which shrinks as awareness rises — people who have formed a view in wealthy countries lean negative on jobs and inequality, while in many middle-income countries a third or more have not formed one. Inside a single country, the people who have heard a lot are disproportionately younger, better educated and better off, and those groups are more relaxed about AI in general. This reading is the page's own synthesis of Pew's findings, not a claim Pew makes.
South Korea shows how far apart the questions can drift. Koreans are the most likely of all 37 to expect AI to widen inequality (54%) and tied for the most likely to expect job losses (76%) — yet only 18% are primarily concerned about it, one of the lowest shares anywhere; 61% say they are equally concerned and excited and 21% primarily excited. Expecting disruption and fearing it are not the same attitude.
The young are catching up — to their elders' worry
Everywhere Pew could measure it, adults under 35 are more likely than those 50 and older to have heard a lot about AI — in Spain, 67% vs. 16%. In the U.S. the share of 18-to-34-year-olds who have heard a lot rose from 31% in 2022 to 59%; among those 50 and older, from 22% to 40%. In 15 countries the older group is the more concerned one. But in several places concern among the young is rising faster, and in two places the gap is gone: since 2024 the share of Americans 18 to 34 who are more concerned than excited has climbed from 40% to 55%, level with those 50 and older (53%), and in Sweden the young (45%) now edge past the old (43%).
| Country | 18–34 · 2025 | 18–34 · 2026 | Change | 50+ · 2025 | 50+ · 2026 |
|---|---|---|---|---|---|
| U.S.† | 47 | 55 | +8 | — | 53 |
| Sweden | 24 | 45 | +21 | 36 | 43 |
| Poland | 24 | 37 | +13 | 44 | 52 |
| Brazil | 32 | 44 | +12 | 60 | 58 |
| Japan | 14 | 25 | +11 | 38 | 34 |
| Australia | 42 | 51 | +9 | 57 | 58 |
Pew shows only countries where the rise among 18-to-34-year-olds is statistically significant. † U.S. 2025 value for 18–34 derived from Pew's labeled 8-point change; Pew's chart does not label the U.S. 50+ value for 2025. For comparison, 2021: 18–34 at 32%, 50+ at 42%.
The thread
The February survey showed that Americans are skeptical; the global release shows they are unusually skeptical. They are among the most likely anywhere to expect lost jobs, a wider wealth gap and more to fear than to welcome — and the rise in concern among young Americans is visible in other rich countries too. The countries that look calmest are not, for the most part, more optimistic. They are less certain. As awareness spreads, the open question is which way the undecided will break.