Few technologies have captured the public imagination as quickly—and as intensely—as artificial intelligence (AI). Pew Research Center data (2025) paint a striking picture: AI is on nearly everyone’s radar, but not all awareness is created equal. Ninety-five percent of U.S. adults have heard at least a little about AI, and the portion of adults who report that they have heard or read “a lot” about AI has nearly doubled since 2022—from 26 percent to 47 percent. Still, gaps persist beneath these impressive numbers.

• Education is associated with a gap in knowledge about AI: six in ten adults with postgraduate degrees report knowing a lot about AI, compared with just 38% of those with only a high school diploma.
• Racial and ethnic differences exist as well—roughly two-thirds of Asian Americans say they’ve heard or read a lot about AI, compared with around 50% or fewer of Black, Hispanic, and White adults.
• When it comes to AI use, age is associated with how often people engage with AI. While about one-third of adults under 30 use AI at least several times a day, more than half of Americans over 65 use it less than a few times a week.

Yet despite these gaps, Americans across the board agree on one thing: understanding AI matters. Younger, well-educated individuals are slightly more likely to say it is extremely important, but large majorities recognize at least a need for AI literacy.

AI is everywhere, but not everyone is on equal footing. Some understand it deeply, others barely at all. The real challenge isn’t access to AI but access to the knowledge needed to engage with it meaningfully. Are we ensuring that everyone has a fair shot at mastering, shaping, and thriving alongside AI?

Fault Lines of AI Discourse

I currently work as a research fellow at the Center for Trustworthy AI at Seoul National University in South Korea. My first role was on the organizing team of the Seoul AI Policy Conference (SAIPCON), an international forum where experts from around the world gather to discuss global AI governance and related policy challenges. The conference is open to anyone in the general public with an interest in AI.

This year’s theme was “Fault Lines of AI Governance,” a phrase that points to the tensions that arise as we develop AI. Fault lines appear where technical capability, economic interests, societal norms, legal frameworks, and ethical principles clash. For example, generative AI is advancing faster than regulation can keep up. Algorithms promise efficiency, yet can reinforce inequity. AI systems that are now widely adopted can collide with local cultures, raising questions about whose values shape “standards.” Data-hungry models strain privacy, and the line between automation and human agency gets blurred.         

As I listened to scholars, policymakers, industry leaders, and legal professionals discuss these topics, my thoughts drifted to different questions: How close are experts’ debates to the general public’s understanding of AI? And more urgently, how distant are experts from the lived experiences of those who are excluded from these conversations? What does AI governance mean if it fails to address the concerns of those already pushed to the margins?

Perhaps the true fault lines lie in the very divides that leave lay publics out of these discourses—where AI, the most transformative technology to date, can also become the most disruptive or exclusionary. Unless the voices of those outside expert circles are included, the divides will only deepen. Inclusive AI governance is not optional—everyone, whether more or less engaged, deserves a seat at the table.

Thirst Meets Apathy — At the Same Edge of Vulnerability

In my previous article, I shared a story about my father—now in his 70s—who has become an enthusiastic ChatGPT user. The moment a new technology appears, curiosity takes over. The moment he gets his hands on it, his inner learner wakes up. As I watched my father try out ChatGPT, it hit me: motivation doesn’t automatically turn into know-how.

Not long ago, my father took part in a church event commemorating his many years of service. A pamphlet was produced for the occasion. When my family looked through it, we immediately noticed something was wrong. The ID photo under my father’s name had been replaced by someone else’s—a small mistake, perhaps, but one that felt disproportionately disappointing.

On the way home, my father admitted what had happened. He had uploaded a photograph of himself from 10 years ago and asked ChatGPT to “make him look seventy.” The face had been aged—but it was not his. (Why he failed to notice this remains the more troubling question.)

My mother is the polar opposite of my father. She isn’t curious about new technologies. She doesn’t chase updates, features, or tools—and she gets along just fine without them (even though, of course, those technologies are woven far more deeply into her everyday life than she realizes). To her, AI is something she has heard of and it’s probably “good to know about someday,” as the above Pew Research findings described. But that’s where it ends. She has almost no real engagement with it, and no desire to start.

So, when she recently told me—with absolute conviction—that a famous celebrity had just given birth, I was caught off guard. I hadn’t heard a single mention of it, so I instinctively questioned it. As always, I gently suggested she might have come across fake news. But she doubled down, and then came the “proof”: a social media photo circulating online of the celebrity holding a newborn baby. One glance told me the image was AI-generated. She, however, saw no difference. For her, the picture was real—it was evidence. And in that moment, I realized that her confidence wasn’t rooted in knowledge, but in trust—trust in the image, trust in the feed that delivered it, trust in the world she experiences through screens.

In that moment, something clicked. These two parental profiles—hyper-motivated and barely interested—stood on opposite ends of the digital spectrum, yet they met a shared point of vulnerability to AI use.

Fueling Familiarity on the Surface, Breeding Overconfidence Underneath

Building on this very point, my new interview study, AI-Enabled Digital Literacy Support: Embracing Readiness, Confronting Vulnerabilities, explores how individuals in the midst of developing digital literacy make sense of AI and use it. The study looks closely at those in the middle of the learning curve: people who are neither fully confident nor disengaged, but navigating new AI tools while trying to learn about this new technology. The focus is thus on the transitional state of learning, where curiosity, uncertainty, and vulnerability coexist.

The interviews were conducted through the community outreach network of Marylanders Online, a statewide digital equity initiative at the University of Maryland. This study was a collaboration with Dr. Mega Subramaniam (Associate Dean for Faculty at the University of Maryland College of Information) and graduate researchers Uhjin Sim and Antariksa Akhmadi from the same college. In total, 22 Maryland residents between the ages of 25 and 44 participated. Many had used digital literacy support (DLS) services offered by public institutions such as libraries, schools, community centers, nonprofits, and advocacy groups, typically several times a month. Overall, they saw these services as moderately effective and important in meeting their digital literacy needs. We met each participant for a 30-minute Zoom interview, where we asked about their past experiences with traditional, human-run digital literacy services, their exposure to AI, and, more importantly, their thoughts on newly emerging AI-enabled digital literacy support (hereafter ‘AI-DLS’). AI-DLS includes things like chatbots that offer 24/7 tech help or intelligent tutoring systems that recommend learning resources tailored to individual needs. Many of these tools can carry on conversations using natural language processing, respond proactively to users’ questions, and adapt their feedback over time based on people’s digital behavior.

Here are three questions we explored.

First, when people look for digital literacy help, how do they feel about AI tools compared to getting support from real humans? Interviewees said that human-led digital literacy services were generally helpful—especially when they needed feedback or hands-on support to fix tech issues. But their experiences also depended a lot on the staff’s expertise, and they often ran into practical barriers like limited service hours, long travel distances, and language hurdles. By comparison, people found AI-enabled services to be simpler and far more wide-ranging. They loved getting instant guidance, the information they needed right away, without the stress or awkwardness that can come with talking to a real person. Some even described AI tools as a kind of virtual companion—something that’s always available, boosts their mood, and can speak their preferred language, which they said is often hard to find in English-dominant, in-person services. While interviewees didn’t report many problems with AI itself, they did point to infrastructure issues that can hamper AI use, like slow or unreliable internet, outdated devices, and weak connectivity in rural areas. They felt that improved connectivity and devices would help more people take advantage of AI. Overall, they believed AI-DLS works best for people who already feel comfortable with technology, whereas older adults, tech-averse users, or anyone unsure about what to ask—or how AI even works—might find it harder to use.

Second, how much do people trust AI-DLS—and do they see any risks to using AI? The answer to this question turned out to be surprisingly straightforward. Across interviews, people described a remarkably high level of trust in AI-DLS. Many said they had never encountered anything harmful or misleading from AI-DLS, and they framed AI as reliable, consistent, and even transparent. Several participants talked about AI almost like an information hub—“a mix of different sites coming together”—that simply delivers what they need. What stood out even more was how unconditional this trust often felt. Many interviewees said they had no reason to doubt AI as “tools,” and a few even expressed near-total confidence, trusting it “99%” as long as they asked clear, reasonable questions of it. When ethical concerns like data security or safety did come up, they tended to pass quickly—often dismissed with a shrug and a sense that it wasn’t a personal concern. Some participants even framed data collection in the course of interacting with AI as inevitably useful, believing it helps AI better understand their needs and tailor its support. In short, people trusted AI because it worked well for them and because they saw little reason to question its output. Data leaks or privacy breaches felt distant, abstract, and outweighed by the convenience and clarity AI tools provided.

Finally, what would it take for AI-DLS to feel more approachable for more people? Across the board, interviewees were optimistic: they saw AI as something that could make DLS more reliable and tailored to people’s needs. But they also pointed out three big areas where extra effort is needed to make AI-DLS work more fairly for everyone in terms of interface, implementation, and infrastructure.

• What people wanted most was simplicity: AI interfaces with bigger text, intuitive design, and instructions that felt easy to follow, especially for senior users. They kept coming back to the same point—when it comes to fixing internet issues or adjusting device settings, videos and visuals help far more than the long, text-heavy responses still typical of many large language model-based tools.
• People wanted AI-DLS to be more adaptive to real user needs. Instead of just answering questions, they imagined tools that are implemented to learn from how people actually use technology—tracking digital habits to offer support at the right moment. And again, language came up: they wanted strong multilingual support, not English-only systems.
• People were clear about one thing: no tool works in isolation. Reliable infrastructure still matters—along with affordable devices, stable networks, and ongoing, human-led support through trusted community anchor institutions. Some comments stood out in particular. People talked about the need for government investment in rural connectivity, lower device costs, and managed service programs—or free resources—to support participation among lower-income users. A few imagined AI-DLS becoming an “built-in, permanent app” available across all devices, not just something you need to download.

From Confidence to Competence: What AI-DLS Should Be

Through this study, we gained an understanding of three key points that matter when it comes to the promise of AI-DLS. While these insights aren’t necessarily generalizable, they still highlight critical considerations that digital literacy practitioners should pay close attention to as AI-DLS continues to evolve in increasingly meaningful ways.

Ensuring Informed and Reflective AI Use

Our interviews suggest that many DLS-seekers accept AI outputs with little questioning. As AI-DLS becomes faster, efficient, and seemingly neutral, trust comes easily—sometimes too easily. This uncritical reliance is reshaping the role of digital literacy support.

In this context, teaching people how to prompt AI tools is no longer enough. What is increasingly needed is critical AI literacy: the ability to think with AI, recognize its limitations, and know when human judgment should take precedence. The appeal of AI-DLS also comes with a trade-off. Concerns about privacy, bias, and security often fade into the background, revealing a growing willingness to exchange caution for convenience. This makes clear that AI literacy must extend beyond technical skills to include discernment and ethical awareness.

One promising example, mentioned earlier, is Marylanders Online, which offers self-directed AI literacy resources organized around everyday use—such as AI tools for productivity and learning, as well as guidance on AI privacy and safety. Rather than focusing only on tools, the curricula address engaging questions like “AI Algorithms: How Well Do They Know You?”, “AI Chatbots: Who’s Behind the Screen?”, “Automation and AI: Should We Be Excited or Concerned?,” prompting learners to think with greater awareness about how AI works and what it asks of them.

Honoring the Human Care That AI Can’t Replace

As our interviewees talked about their experiences with AI-DLS, a pattern emerged: AI-DLS seems to balance the comfort of “warm experts,” who feel friendly and familiar, with the efficiency of “cold experts,” such as tech specialists or helpdesks (Hänninen & Taipale, 2025). This hybrid quality makes AI-DLS feel both emotionally safe and practically effective, helping explain its growing appeal as a complement to human-led services.

Yet the very relational appeal does not render human-driven DLS obsolete. Instead, it clarifies what remains uniquely human in digital literacy work. As AI-DLS becomes more conversational and companion-like, human staff continue to play a critical role as sources of emotional grounding and ethical orientation. While AI can generate warm, reassuring language, what ultimately matters is not tone alone but contextually attuned response—language shaped by an understanding of lived circumstances and unspoken concerns.

In this sense, DLS staff function less as information providers and more as social anchors. They offer care, equity work, and discernment that cannot be automated or simulated. AI systems may sound empathetic, but they do not empathize; they reflect emotion without sharing it. Human supporters demonstrate that genuine connection and responsibility cannot be fully delegated to AI.

Rethinking the AI Divide in Digital Literacy Support

Uneven access to devices and broadband continues to widen the AI divide—especially in rural and low-income communities. Still for many learners, the problem is not whether AI-DLS works, but whether the learners can access these tools at all. Without addressing these basic gaps, AI risks benefiting those already well-positioned, while leaving behind the very groups it promises to support.

This is where AI equity matters. Community anchor organizations see what AI cannot: local realities, affordability barriers, and disability-related needs. Their role is not just to introduce AI-DLS, but to guide how these tools are used—stepping in when technology amplifies inequality rather than reducing it.

Closing the AI divide, then, is not only about smarter systems. It is about people who ensure that AI serves learning fairly, responsibly, and inclusively.

Dr. Yeweon Kim is a postdoctoral researcher at the Center for Trustworthy AI at Seoul National University and a Benton Opportunity Fund Fellow. Her research explores the societal impacts of information and communication technologies at individual and community levels, with a focus on digital empowerment, equity, and ethics. 

Also by Dr Kim: The Invisible Divide: Why Bridging the Digital Divide Begins with the Mind, Not the Metrics

References

Hänninen, R., & Taipale, S. (2025). Warm experts among us: Conceptualising the challenges of informal digital support for older adults. New Media & Society, 14614448251385087. https://journals.sagepub.com/doi/10.1177/14614448251385087

Kim, Y. W., Sim, U., Akhmadi, A., & Subramaniam, M. (2026). AI-enabled digital literacy support: Embracing readiness, confronting vulnerabilities. The Journal of Community Informatics, 22(2). https://doi.org/10.15353/joci.v22i2.6499

Pew Research Center. (2025, September 17). How Americans view AI and its impact on people and society. https://www.pewresearch.org/wp-content/uploads/sites/20/2025/09/PS_2025.9.15_AI-and-its-impact_report.pdf

Marylanders Online. (n.d.). https://marylandersonline.umd.edu/home/

Seoul AI Policy Conference (SAIPCON). (2025). https://www.saipcon.com/

The Benton Institute for Broadband & Society is a non-profit organization dedicated to ensuring that all people in the U.S. have access to competitive, High-Performance Broadband regardless of where they live or who they are. We believe communication policy - rooted in the values of access, equity, and diversity - has the power to deliver new opportunities and strengthen communities.

© Benton Institute for Broadband & Society 2026. Redistribution of this email publication - both internally and externally - is encouraged if it includes this copyright statement.

For subscribe/unsubscribe info, please email headlinesATbentonDOTorg

Kevin Taglang
Executive Editor, Communications-related Headlines
Benton Institute
for Broadband & Society
1041 Ridge Rd, Unit 214
Wilmette, IL 60091
847-220-4531
headlines AT benton DOT org

Share this edition:

Broadband Delivers Opportunities and Strengthens Communities