A 10,000-Person Study Found That AI Awareness Tracks Wealth and Education
- David Borish
- 3 days ago
- 5 min read
The average American adult uses AI every day without knowing it. Recommendation engines sort what they watch. Chatbots handle their customer service calls. Email filters quietly triage their inboxes. Yet according to a new peer-reviewed study from Hong Kong Baptist University, whether someone can actually identify these systems depends significantly on their income and education level.
, analyzed responses from 10,087 U.S. adults surveyed by the Pew Research Center in December 2022. Participants were tested on their ability to identify AI-enabled tools in six common everyday scenarios, including online shopping, email services, and customer support. The average score was 3.72 out of 6. But that average masked a pattern tied directly to socioeconomic position.
What the Numbers Show
Individuals with higher education levels scored significantly higher on AI awareness, with education producing a standardized coefficient of .27 in the regression model. Household income produced a separate, independent effect (.20). Both held up after controlling for age, gender, ethnicity, and geography.
The gaps did not stop at awareness. Higher-income, higher-education respondents were also more likely to report using AI and to rate themselves as more familiar with it. Education was a stronger predictor of usage than income, a finding the authors note is consistent with research on internet usage gaps going back two decades.
The familiarity result is worth pausing on. Researchers measured it not as factual knowledge but as perceived familiarity, meaning how much people felt they had heard or read about AI. Higher-SES respondents rated themselves more familiar on a three-point scale, even after controlling for all demographic variables. This matters because the study then found that perceived familiarity was a stronger predictor of AI awareness than actual usage behavior. Knowing about AI through media consumption and conversation, even without directly using it, appears to build more awareness than simply encountering it through an app.
The Mechanism Behind the Gap
The researchers tested whether usage and familiarity could explain why socioeconomic status predicts awareness, rather than simply correlating with it. The mediation analysis confirmed both pathways. Higher income and education led to more AI use, which in turn raised awareness. They also led to greater perceived familiarity with AI, which had an even stronger effect on awareness.
This matters for how the problem gets framed. The most obvious response to an awareness gap might be to provide access: make AI tools available to lower-income populations and the gap will close. The study's findings suggest that logic is insufficient. Familiarity built through indirect exposure, reading about AI, discussing it, encountering coverage of it, does more work than direct use. Someone who uses an AI-powered tool without knowing it is AI does not close their awareness gap through that use alone.
The authors describe this through the concept of "experience technology," a framework from earlier internet research. People learn about technology through repeated engagement and reflection on that engagement. Passive or incidental use of AI, the kind that happens when you ask Alexa something or get a product recommendation on Amazon, does not produce that reflective engagement. It builds familiarity only when paired with some broader context for understanding what is happening.
Who Gets Left Behind, and Why It Matters
AI awareness is not simply a curiosity metric. The study's authors frame it as a new layer of digital inequality, sitting above the older categories of access, skills, and outcomes. Unlike a computer or a smartphone, which users can see and consciously choose to learn, AI frequently operates as invisible infrastructure. Streaming services, search engines, and social platforms run AI continuously without surfacing it. Someone who cannot identify these systems is also less equipped to question them, challenge algorithmic bias, or protect themselves from data practices they have not knowingly consented to.
The sample demographics are instructive. More than 20 percent of respondents had household incomes below $30,000. About 35 percent had not completed a four-year college degree. These populations scored lower on AI awareness across the board. They reported using AI less frequently and felt less familiar with it. The study found that Black and Hispanic respondents also scored lower on AI awareness, though these patterns largely followed the socioeconomic distribution rather than ethnicity alone.
Gender differences appeared throughout the results. Male respondents reported higher AI usage, familiarity, and awareness than female respondents, consistent with research on gender and technology engagement more broadly. Age followed expected lines as well: younger respondents scored higher on all three measures, likely reflecting the degree to which younger generations already live within platforms that embed AI heavily.
The Literacy Implications
The study connects its findings to a broader argument about AI literacy, which researchers define as the ability to identify, use, and critically evaluate AI and its associated risks and responsibilities. Awareness, the ability to recognize that AI is present at all, is the precondition for everything else. You cannot critically evaluate a system you do not know is operating.
This creates a compounding problem. Individuals who are unaware of AI are also, the authors note, less likely to seek it out, less likely to develop the skills to use it effectively, and less likely to benefit from the productivity gains it offers. Meanwhile, higher-income and higher-educated populations, who already have more resources, are gaining AI capabilities faster. The gap is not static.
The authors recommend integrating basic AI concepts into educational curricula as a way to build familiarity among populations with limited direct exposure. Workshops, public campaigns, and accessible online resources are cited as supporting mechanisms. The key insight is that building familiarity, not just access, needs to be the target.
Limitations Worth Noting
The dataset comes from December 2022, before the public proliferation of generative AI tools like ChatGPT, which launched the same month the survey was fielded. It is plausible that the subsequent two-plus years of AI coverage and tool adoption have shifted some of these patterns, though the structural dynamics linking income and education to awareness are unlikely to have reversed. The study also relies on cross-sectional data, meaning causality cannot be firmly established. The authors acknowledge that awareness might also drive usage, not just the reverse.
The AI awareness measure itself is specific: it tests recognition of AI in predefined everyday scenarios, not a broader understanding of capabilities, risks, or implications. Deeper forms of AI literacy remain largely unmeasured in the survey literature.
Despite these constraints, the study is among the largest and most methodologically rigorous examinations of AI awareness disparities in the United States. Its core finding, that socioeconomic position shapes not just whether people use AI but whether they know they are using it, has direct implications for how policymakers and educators approach AI literacy initiatives.
The Gap AI Is Quietly Widening
The technology conversation tends to focus on what AI can do. This study redirects attention to what people understand about what AI is already doing to them. Awareness is not glamorous, but it is foundational. People who cannot identify the AI systems shaping their information, their job applications, their loan decisions, and their healthcare recommendations are at a structural disadvantage compared to those who can.
That disadvantage tracks, with considerable precision, along the income and education lines that define economic opportunity in America more broadly. Whether the next wave of AI policy and education addresses that reality, or proceeds as if equal access to tools is the same as equal footing, is a question with consequences that will compound over time.
Click image to read the previous article
Comments