The Age of AI: July 2025

Summary

Depending on who you ask, artificial intelligence (AI) tools either pose an existential risk to humanity, or will supercharge productivity and economic growth to unlock unforeseen levels of flourishing. Sam Altman, CEO of OpenAI, said in January, 2025 that the tools his company is working on may soon require a rewriting of the social contract, and fundamental rewiring of the understanding of human labor.

Given the stakes of this new technology, researchers should investigate how people are reacting to the excitement — and anxiety — of widespread AI advances. It is important to hear from a representative sample of Americans about how the technology is impacting their lives, not just from startup investors or AI naysayers. In June, Verasight launched a project to track these impacts on a monthly basis. In this first poll, we interviewed a nationally representative sample of 1,500 adults on their use of AI tools at work, at school, and in their personal lives. Data were collected from June 19–June 24, 2025. This is the first in a set of regularly updating surveys on AI adoption in the US.

Overall, we find wide adoption of AI across domains. Sixty-one percent of respondents to our survey said they use an AI tool at work or in their personal life in the last month. Twenty-three percent said they use some sort of AI tool at least once a day, and a total of 43 percent say they use an AI tool at least once a week. This accounts for over 140 million weekly users.

In probing how Americans use and feel about AI, we uncover three big themes. First is an underlying anxiety among workers, especially white-collar workers, about AI disruption to their careers. Roughly 15 percent of workers say AI could do "all" or "most" of their jobs) and roughly one in ten recently terminated workers say AI either replaced them or "reduced the need for" their role. Blue-collar workers fared better (6 percent) when compared to office workers (12 percent), but were not immune to AI. Across our full sample, 43 percent of Americans said they were worried about AI taking jobs away from humans.

The second theme is the deep penetration of AI tools in education, and the unpreparedness of policy response to technological change, particularly large language models. In our survey, over half (53 percent) of students report using an AI tool that has not been sanctioned by school administrators on school work. Students guess that roughly two-thirds of their peers are using AI on school work.

And finally, in personal life, AI’s footprint is expansive but not yet transformative. Common uses include drafting messages, seeking recommendations, and organizing schedules, yet 75 percent of users say AI has not fully replaced tasks they once did manually. Outside of the office, people see AI as presenting few uses. For non-users, the biggest barriers are perceived lack of need (51 percent) and distrust of AI (35 percent).

Overall, Americans are navigating AI with experimentation, and caution. The technology’s social contract is actively under negotiation, with policy, education, and workforce development lagging behind consumer curiosity and corporate hype. This survey underscores the need for clear frameworks from governments, businesses, and parents on acceptable uses of AI. The most common feeling about AI is not optimism (40 percent) or excitement (33 percent), but anxiety (53 percent).

‍

‍

1. Use and awareness

Brand recognition and usage

Our survey shows that ChatGPT by OpenAI has a small first-mover advantage in the AI market. Figure 1 displays the percentage of all respondents who say they have heard of each of eight leading LLM providers, in column 2, and the percentage of people who have heard of at least one model that have used each, in column 3. While ChatGPT is the most recognized, at 85 percent brand awareness, Google Gemini comes in a close second at 76 percent. Among all other model companies, only Microsoft Copilot has a brand awareness among 50 percent.

Figure 1

‍

Overall, 61 percent of Americans report using, and 39 report not using an AI tool at work or in their personal life in the last month.

In terms of usage, ChatGPT is also top-dog. Our survey shows it is used by 41 percent of all Americans. Google Gemini is a close second, with 33 percent of Americans reporting use over the last month.

No other company cracks 20 percent use, though Microsoft comes close at 18 percent. Despite this, some companies have intense bases of users. For example, Claude and Perplexity both rank rather low in terms of awareness, but match other well-known models (such as Elon Musk's Grok) in usage. These companies have leaned heavily into two applications of their tools — coding and improved search/global knowledge retrieval — and show the utility in the AI space of finding a niche.

Meta's Llama models rank lowest for combined awareness and usage, with just 12 percent of Americans knowing about the model, and 1 percent using it. Llama ranks lower on awareness even than DeepSeek, the model from an AI startup in China that briefly surpassed the performance of American models in early 2025.

Age gradient

Age is the primary determinant of brand familiarity (Figure 2). Ninety-four percent of 18- to 29-year-olds are aware of ChatGPT, compared with 83 percent of seniors 65+. The youngest cohort is also twice as likely as seniors to recognize insurgent brands like Grok (42 percent vs. 22 percent) and Perplexity (17 percent vs. 4 percent). Income and education amplify but do not supplant age dynamics; college-educated adults and households earning $100 k+ display five- to ten-point boosts in awareness across brands.

Young people are also much likelier to use AI models, though use is not as ubiquitous as popular narratives convey. Forty-nine percent of respondents ages 18 through 29 say they used ChatGPT over the last month, for example, whereas just 40% of Gen X Americans use it and penetration is just 25% among Seniors. Young people are twice as likely to say they use xAI’s Grok model compared to elderly, though use remains low at 6% of all Americans.

Figure 2

‍

Racial/ethnic differences are modest once socioeconomic status is controlled, suggesting that media exposure and professional contexts are driving awareness. One interesting finding is that, aside from ChatGPT, usage tends to increase among Millennial and Generation X Americans compared to Gen Z. This is likely because of the penetration of these tools in corporate office products, such as Gmail and Outlook.

Non-users

For the 39 percent who abstain from AI, perceived need dwarfs all other barriers; 51 percent “don’t need it,” far outstripping privacy (31 percent) and trust (35 percent) concerns. This finding challenges narratives that assume fear alone suppresses adoption. Many Americans simply have not located a value proposition.

Figure 3

‍

Age patterns invert here. Older respondents cite unfamiliarity and learning barriers (“I don’t know how to use it”) at double the rate of Gen Z—34 percent vs. 14 percent — whereas environmental worries resonate more with Gen Z (41 percent) than any other cohort.

American AI typology

Based on all this information, we have sorted Americans into one of five groups, for tracking purposes

• Early Adopter: People who know multiple tools, especially from niche companies, and use them multiple times per da
• AI Normie: People who use one or two of the most popular tools, mostly for asking questions, at at least a weekly rat
• Behind The Curve: People who report using AI at a low frequency
• Generic Non-user: People who report not using AI tools, but because of mundane reasons (such as cost, not needing it or knowing how
• AI Skeptic: People who report not using AI tools, and who have one or several serious reservations about them (they are low-trust or are concerned about impacts to privacy, society, or the environment)

In our first survey, the breakdown for each group is visualized below:

Figure 4

‍

2. AI at work

Adoption

Since we had a large sample, we were able to filter it to people who are employed or temporarily unemployed (about 67 percent of our overall sample, or 1,000 respondents).

A significant minority of users across industries report using AI at least weekly. Thirteen percent say they use AI tools several times per day, eight percent once daily, and 14 percent "a few times a week" — for a total of 35 percent weekly+ user base.

Most reported AI use at work is relatively mundane (see chart). Most people use AI to write for them, including in emails and reports (30 percent), or summarize text they are assigned to read (19 percent). Other use AI to manage their tasks (19 percent), including automation of repetitive work (14 percent). A sizable proportion of workers also report using AI for creative work, such as data analysis (19 percent) brainstorming (19 percent) or creating visuals for presentations (14 percent).

Figure 5

‍

The most common tool used by workers is the conversational AIs, such as ChatGPT. Thirty-two percent of workers report using these tools in their jobs. Next, general productivity tools embedded in existing software — such as Google Docs and Microsoft office — are used by 26 percent of workers. In a near third-place, writing tools such as Grammarly are used by 17 percent.

Comparatively few workers use other AI systems, such as for coding (6 percent) and photo or video editing (12 percent combined).

It is much more common for young people (31 percent) to use conversational AIs at work, compared to 20 percent for workers above the age of 65.

Based on the industry people report working in, we are also able to look at AI usage among white-collar and blue-collar workers. Overall, 68 percent of white-collar workers report using AI with at least some frequency at work, including 28 percent who use it at least once a day. In contrast, Just 35 percent of blue-collar workers use AI at work, and just 10 percent do so daily.

White-collar workers’ greater adoption stems partly from frictionless integration: office software now auto-suggests AI writing or data-cleanup features. The general availability of models on easy-to-use websites also presents an opportunity for AI companies. In contrast, blue-collar tasks often require physical work, which AI has not yet replaced.

Across multiple questions, we find a pre-existing deep penetration of AI tools into the typical office work today. For comparison, it is likely that AI use is now as or more common than the landline desk phone, a hallmark of the 21st century office job.

Figure 6

‍

Anxiety and job security

We also find that white-collar workers have higher levels of anxiety about AI replacing them, though broken down by income, this is not who you’d expect.

As mentioned in the summary of this report, many people are worried about the impacts that AI tools will have on the job market. Fully 66 percent of Americans (not just workers) think that AI will lead to fewer job opportunities in the future. That is only slightly higher for young people (70 percent) than older workers (64 percent), and is slightly higher among lower-income workers (70 percent) than the wealthy (62 percent) and those with college degrees (66 percent). Regardless of their demographics, Americans are worried about losing their jobs to AI.

Among people who lost a job, a sizable proportion say that AI either replaced them outright (3 percent) or reduced the need for their role (7 percent). This was more common for young people (14 percent total), lower-income workers (14 percent), and white-collar workers (12 percent).

These statistics challenge the predominant narrative of AI reducing demand for high-paying, junior white collar positions, such as in finance roles. Our survey finds that lower-rung, low-paid positions in offices — such as clerks and administrative assistance — are the most threatened positions in a post-AI world. Although only a small percentage of workers report being made redundant by AI, an increase in the employment rate of even a few percentage points could have economic impacts, and these stories have been covered frequently and negatively by the media.

Ironically, our survey reveals that executive- and manager-level employees are the most likely to say AI could replace most or all of their current job, when compared to junior employees or service workers. Fully one-quarter of executives, including CEOs and presidents, told us they thought AI systems could replace all or most of their jobs, compared to 7 percent for service workers:

Figure 7

‍

Future and upskilling

Respondents were also asked what skills they thought were necessary to be a successful worker. Many workers cited "soft" skills: the most common response was "interpersonal skills, such as getting along with people and resolving conflicts" — which 79 percent of adults said was either "extremely" or "very" important to success. People also frequently picked skills such as writing and communication (77 percent), management (69 percent) and basic compute skills (74 percent).

Unfortunately for human workers, these are all things AI systems are increasingly skilled at. An AI is perfectly amicable if told to be, and increasingly skilled at writing and even creating presentations. If work is conducted on a computer, most entry-level skills, while important, can be replaced with AI. Only 30 percent of workers said that the "ability to integrate AI" is important to success as a worker.

Many workers also thought the "hard" skills required for trade jobs were important. About 60 percent of respondents each identified manual labor or working with machinery as crucial for success in the workplace. These jobs have also been replaced at lower rates, according to our poll.

Overall 21 percent of workers, including 24 percent of young workers and 29 percent of office workers, reported feeling pressure to use AI to keep up with others in their field. The irony is that by learning how to use AI in their work, many white-collar workers are increasingly the probability that AI replaces them.

Above all else, 80 percent cited critical thinking as key to success.

3. AI at school

Policy gaps

Despite the intensity with which students are using AI, policy in schools is broadly lacking. Our survey also shows many students are breaking the rules administrators have set for them.

Only 8 percent of students say AI is fully approved and integrated in their schools, while 35 percent note conditional approval requiring teacher sign-off. One-fifth report using tools that are outright prohibited by their school, and 18 percent are unsure of any policy. (We can be reasonably sure a large proportion in the final category are using AI tools despite this uncertainty).

Many students report using tools in violation of school rules, and with some regularity. Eight percent of the student sample said they use AI tools that violate school rules "very often," ten percent said they do so "often," and an additional 34 percent said they do so "occasionally" or "rarely". Overall, a majority of students say they use AI in unapproved ways. One-fifth of students responded that they use AI to answer test questions.

Five percent of our respondents (20 percent of those ages 18-29) indicated they were currently enrolled in school of some sort. Three percent of them are highschoolers, 19 percent are enrolled in community college, 61 percent are pursuing a bachelor's degree, 14 percent are pursuing a tertiary degree, and 4 percent are enrolled in a trade program.

While this is a small sample of AI users (50 people; not all students report using AI), our survey indicates this group is the most actively engaged with AI tools. Overall 88 percent say they have used AI tools in the last month, including 57 percent who use AI tools weekly.

Students report using AI for the following reasons

• Brainstorming ideas: 50 percent
• Studying/exam prep: 43 percent
• Grammar/writing improvement: 42 percent
• Math/science problem-solving: 39 percent
• Summarizing lectures: 38 percent

4. AI in personal lives

Nearly half (48 percent) of Americans report using AI in their personal lives. At home, AI use mostly revolves around digital communication. If you've received a message from one of your friends over the last year that was wrapped in quotation marks, or began the text with "here would be a good thing to send to your friend", then you know what we're talking about.

Figure 8

‍

Some Americans use AI only at work, and not at home. But of the 61 percent who use AI at all, just 13 percent of respondents said they don't use AI in their personal lives. It seems like if you use AI for one part of your life, you find applications elsewhere.

Personal AI use centers around writing, just like in the workplace. Respondents report using AI tools mostly to write or review messages and social media posts, or get recommendations for entertainment, organization, and health/fitness.

Household labor

Because one of the promises of AI is to reduce the amount of time humans spend on mundane or repetitive work, we asked respondents to tell us whether AI systems now do work for them that they used to do themselves. Twenty-five percent of respondents said yes, and provided their own accounts of how they use AI.

By far the most common application was writing, especially email. Respondents said they used AI to "write more difficult emails" and "construct professional emails to people I can’t stand". Many respondents also said they use AI to do research for them (including "researching laws") and write cover letters and job applications.

Very few respondents indicated AI helps them with any meaningful work at home. That is unsurprising, given the limits of AI tech today, but notable given the nature of household work. Some respondents said AI writes recipes for them, or makes grocery lists.

Despite the promise of this technology, it is mostly limited to applications in learning and the workplace.

Anxiety about societal consequences

In addition to our earlier discussion of AI-induced anxiety in the workplace, it is worth dwelling briefly on overall anxiety about AI's place in society.

Figure 9

‍

With 56 percent “extremely/very” concerned, misinformation in AI output ("hallucinations") tops the worry list. People are equally worried about others using AI in cyberattacks, as well as a "loss of human creativity and interaction" resulting from artificial intelligence.

The next most common worry is that AI systems will escape human oversight (54 percent), and that humans will lose jobs to AIs (49 percent). Recall that 66 percent predict fewer opportunities.

Younger workers paradoxically express both higher optimism about AI’s capabilities and heightened fear of replacement — a cognitive dissonance even more notable based on the frequency with which young people say AIs are replacing them, and that they are adopting AI in their work and school lives.

Environmental impacts of AI data centers rank lower overall (37 percent extremely/very concerned) but surge among Gen Z (49 percent). Ethical data sourcing and concerns about bias garners similar demographic skews.

5. Conclusion

Artificial intelligence in mid-2025 is far from mature, the impacts of which can be seen all over our data. There is a paradox in how Americans use AI today: They are very worried about the long-term consequences of AI tools, but are increasingly willing to hand over their time, skills, information and even competitive advantages to them.

Workplace AI use illustrates the paradox. AI already writes emails and summarizes datasets, but the majority of tasks resist full automation, at least for now. White-collar professionals, first to integrate AI, also glimpse their own obsolescence more actively. Blue-collar workers feel safe, for now, but advances in robotics threatens to narrow that gap.

Education mirrors the broader trends in AI and concerns over privacy and de-skilling. Students, digital natives by default, will not remove AI from their study routines any time soon. Institutions must therefore articulate clear, fair, and forward-looking rules.

Looking ahead, three indicators will tell observers how AI is reshaping the social contract

• Governance Pace. Regulation must match technological velocity or risk perpetual catch-up and broad consequences
• Skill Diffusion. Mass-market reskilling, akin to 20th-century literacy drives, will determine whether AI augments or stratifies the workforce
• Trust Infrastructure. Transparent algorithms, auditable training data, and reciprocal accountability between firms and users are prerequisites to avoiding the privacy pitfalls made by surveillance technologies and social media companies in the mid-2010s.