Information Literacy Is No Longer Just an Individual Skill

Based on a talk for the New Jersey State Library's national conference on information literacy

A New York Times reporter recently called me with a question about AI and elections. She wanted to get my take on the political implications of “answer engine optimization” (AEO), also known as generative engine optimization (GEO), the younger sibling of search engine optimization. 

For years, political campaigns, news organizations, public agencies, and advocacy groups have worried about where their information appears in search results: is their webpage among the first links returned by a search engine?

But increasingly, people do not encounter a list of results at all. They ask a question and receive a single, synthesized answer. Knowing that voters rarely click through to the underlying sources, the Missouri state legislative candidate The Times profiled is starting to turn out fact sheets to ensure that his positions on small business show up in Google’s AI-generated summary.

As with most AI questions, GEO is a double-edged sword. Used well, it encourages candidates, organizations, institutions, and knowledge creators to publish better and more accurate information. Used maliciously, bad actors will inject misinformation into the information ecosystem to manipulate what we see.

We now face an equal and opposite challenge to information literacy: information compression. Information compression occurs when thousands of sources are reduced to a single AI-generated synthesis.

We have spent years worrying about information overload. We now face an equal and opposite challenge to information literacy: information compression. Information compression occurs when thousands of sources are reduced to a single AI-generated synthesis.

Given that AI is continuing to change how knowledge is created, as well as how it is found, assembled, and understood, we have to ask: what is the future of information literacy? Who gets heard when thousands of webpages are compressed into a few paragraphs? Who decides which facts, voices, and interpretations survive that compression? And how vulnerable is such compression to manipulation?

These questions point to a transformation in our information economy that goes far beyond elections. 

From information overload to information compression

Information overload predates generative AI. The internet already created more information than any person could possibly read, evaluate, or synthesize. For years, I have repeated the oft-cited IDC statistic about “175 zettabytes by 2025.” It’s such a common stat that it has its own Wikipedia page.

Generative AI intensifies the problem. It makes it cheap and easy to produce an enormous quantity of text, images, audio, and video. Much of that material is useful. Much of it is also low quality, repetitive, misleading, or entirely fabricated.

This is the familiar problem of AI slop (click the link to entertain yourself with John Oliver’s selection of especially funny slop memes). Poor-quality information is generated, published, absorbed into future training data, and used to produce still more poor-quality information. Deepfakes, fabricated news stories, synthetic political propaganda, and machine-generated content make it harder to distinguish reliable information from falsehood.

But information abundance is only half the problem. At the same time that AI is producing more content, it is also reducing more of what we see to a single answer.

Search engines increasingly display AI summaries above conventional results. So when I searched this week to find out:

• Is Mitch McConnell still alive? 

• Why did Lindsey Graham die?

I got a truncated summary of the news.

We live in a world of “zero-click” information where chatbots and search engines now answer questions without requiring users to visit a webpage. 

This is an amplification of the curatorial work that recommender systems previously did (and that editors have always done), namely, deciding what evidence to surface and what to leave out.

A Pew Research Center study found that users clicked on a conventional search result in 15 percent of visits when no AI summary appeared. When an AI summary was present, the rate fell by HALF to 8 percent.

Information abundance forces us to navigate too much. Information compression means we may never see most of it.

Information abundance forces us to navigate too much. Information compression means we may never see most of it.

Together, these changes require us to rethink information literacy.

Information literacy in the AI era has at least three dimensions:

1. How individuals learn.

2. How communities learn together.

3. How institutions listen, learn, and improve.

1. How individuals learn: from searching to questioning

Traditional information literacy focuses primarily on the individual: Can a person find a source, evaluate its credibility, compare it with other sources, and reach a reasoned conclusion?

The central skill is no longer merely finding the right answer. It is asking better questions.

The workflow begins with retrieval:

Search. Evaluate. Compare. Synthesize.

AI increasingly begins at the other end. It offers the synthesis first.

That changes the learner's work. We still need to evaluate sources, check claims, and distinguish fact from fiction. But we must also become skilled at interrogating a generated answer.

What assumptions shaped it? What evidence supports it? Which sources were excluded? What alternative explanations exist? How would the answer change if the question were framed differently?

The central skill is no longer merely finding the right answer. It is asking better questions.

Generative AI can be useful here when treated as a conversation partner. A good conversation with AI can help someone explore an unfamiliar topic, identify missing considerations, challenge a tentative argument, or generate questions for further investigation.

In this sense, AI can resemble the imaginary friends many of us had as children. Those friends did more than entertain us: they helped us rehearse ideas, work through experiences, and imagine possibilities. Generative AI can sometimes serve a similar function for adults: a readily available partner for reflection and inquiry.

But this analogy also reveals the danger. AI systems are designed to respond fluently and agreeably. They often reinforce the user’s assumptions, even when those assumptions are wrong. Researchers call this sycophancy.

Ask AI to challenge your position. Request competing interpretations. Check important claims against independent sources. Use search tools and original documents when factual accuracy matters.

So using AI well requires deliberate habits, and this is where the skills of information literacy taught in libraries and classrooms become essential. Ask AI to challenge your position. Request competing interpretations. Check important claims against independent sources. Use search tools and original documents when factual accuracy matters.

Do not ask the chatbot merely to verify its own answer. 

Most importantly, use AI to strengthen judgment, not replace it.

In one project, our AI for Impact Fellows have been experimenting with an AI coach for people designing public engagement processes. The coach does not tell users what engagement method to choose. Instead, it asks questions: Who needs to participate? What knowledge do they possess? What power will participants have? How will their contributions influence the decision?

The goal is not to automate expertise. It is meant to help people reason more carefully by learning to use AI to have productive conversations with themselves.

2. How communities learn: from individual inquiry to collective intelligence

The most important conversations, however, are not between a person and an AI. They are between people using AI together.

Information literacy has typically been taught as an individual competency. But many of our hardest information problems are collective. Communities must determine what they know, where they disagree, whose knowledge is missing, and how they can act together.

AI can support that process.

Meedan, whose name derives from the Arabic word for a public square, works with communities around the world to identify misinformation and disinformation. Its projects do not simply deploy automated detection tools. They build networks of journalists, civic organizations, fact-checkers, and community members who collectively (with the help of AI) identify rumors, investigate claims, and share reliable information through channels such as WhatsApp.

AI helps groups organize and analyze information. The community supplies context, judgment, trust, and local knowledge.

A different example comes from Innovate Public Schools in California. Parents of children with disabilities often receive individualized education plans that can run dozens of pages and are written in technical language. The process is especially difficult for parents with limited literacy or whose first language is not English.

The library’s role is therefore not limited to teaching patrons how to use a chatbot. Libraries can help communities build collective intelligence using AI.

Working with families, Innovate Public Schools and our AI for Impact Fellows have been developing an AI tool called A-IEP that can translate and summarize these documents and help parents navigate the special-education process. But the most important feature is not the technology itself. It is that parents are helping to decide how the tool should work, what information it should provide, how privacy should be protected, and what families actually need to advocate for their children.

This is community-centered AI: technology designed with and for the people expected to use it.

Libraries are particularly well positioned to support this social and collaborative form of information literacy. They are trusted public spaces where people can learn, deliberate, and solve problems together. They can convene communities to examine how AI systems work, identify local needs, test tools, decide what responsible use should look like, and collectively address:

• What problems matter?

• What information do we need?

• What should the AI do?

• What shouldn't it do?

• How do we know it's working?

The library’s role is therefore not limited to teaching patrons how to use a chatbot. Libraries can help communities build collective intelligence using AI.

3. How institutions learn: from public consultation to radical listening

We have spent decades trying to make individuals more information literate. The next challenge is to make institutions information-literate.

Governments, schools, universities, and other public institutions collect enormous quantities of information. Yet they are often poor at listening, synthesizing experience, learning from implementation, and changing course.

Public consultation illustrates the problem. Institutions may invite comments, hold hearings, or conduct surveys. But listening at scale is expensive and labor-intensive. Thousands of submissions must be read, categorized, compared, and summarized. As a result, consultation often happens infrequently, reaches only a narrow group, or produces information that institutions struggle to use.

AI can reduce some of those barriers.

In New Jersey, AI-assisted engagement helped gather and organize public views about how the state should govern AI. In Bowling Green, Kentucky, a community planning process gathered input from approximately 8,000 people and used AI to help sort and synthesize residents' feedback.

Other projects use generative images to help residents visualize proposed changes to streets and public spaces. Instead of discussing a bike lane, park bench, or redesigned intersection only in abstract terms, people can see alternatives and modify them together.

The use of AI creates more (not less) need for human judgment. AI-generated summaries can omit minority positions, flatten disagreement, or reproduce bias. Institutions must make their methods transparent and allow people to challenge the interpretation of their contributions.

But used carefully, AI can enable a form of radical listening: continuous, large-scale learning from the public's knowledge and experience.

An information-literate institution gathers diverse forms of evidence, explains how knowledge was synthesized, learns from outcomes, and revises decisions.

An information-literate institution does more than publish information. It knows how to ask questions, gather diverse forms of evidence, explain how knowledge was synthesized, learn from outcomes, and revise decisions.

A positive agenda for information literacy

Responding to AI cannot consist solely of warning people about misinformation. We need a positive agenda for building the information environment we want.

First, we need public AI. Just as public libraries exist alongside bookstores and commercial platforms, publicly accountable AI must exist alongside commercial systems. Schools, governments, libraries, and civic organizations need tools that are transparent, open to scrutiny, and designed to serve public purposes.

Second, we need widespread education in the new information literacies and how to use AI for inquiry, conversation, collaboration, and judgment. Training should address risks, but it cannot stop there. People need practical opportunities to learn how to think with these systems without becoming dependent on them.

Third, governments and institutions should require community participation in the design and purchase of AI. Public agencies are among the world’s largest technology buyers. They should not purchase systems affecting communities without involving those communities in determining the tools’ goals, rules, and safeguards.

Finally, we should invest in institutions capable of radical listening. Gathering public input should not be a ceremonial exercise conducted after decisions have effectively been made. Continuous learning should become a core institutional function.

Information literacy is the capacity of communities to build knowledge together and of democratic institutions to listen, learn, and improve.

Why Information Literacy is an Urgent Priority 

The debate over AI is often framed as a choice between embracing technology and rejecting it. That is the wrong choice. 

As Pope Leo wrote in his AI encyclical: “The primary choice is not between a 'yes' or 'no' to technology, but rather between constructing Babel or rebuilding Jerusalem; between a power that claims to dominate the heavens and a people who work together in the presence of God to rebuild the walls of fraternal coexistence.”

The question is whether we can build the skills, communities, and institutions needed to govern our new information environment.

Information literacy is no longer just an individual's ability to determine whether something is true. It is also the capacity of communities to build knowledge together and of democratic institutions to listen, learn, and improve.

That broader form of information literacy is democratic infrastructure.