"Look It Up" Doesn't Mean What It Used to Mean
Across three eras, the same instruction has meant different actions, different information, and different skills. Each transformation traded capabilities we didn't know we'd miss.
Someone asks a question at dinner. You pull out your phone and say, “Let me look that up!” The phrase sounds timeless, as if it has always pointed to the same action. But where you look, how you look, and what happens when you do has transformed three times in recent history, and each transformation appears to be changing how humans think.
“Look it up” once meant walking to a shelf and opening a bound volume. Then it meant typing keywords into a search box. Now it increasingly means asking a conversational AI and receiving an immediate answer. But these aren’t just different methods of finding the same information. Each shift has changed what information people encounter, how they evaluate it, and what cognitive skills they develop in the process of seeking it.
The Shelf Era (1768-1990s): When Looking It Up Meant Knowing Where to Look
The Encyclopædia Britannica first appeared in Edinburgh in 1768 as a three-volume set, marking the beginning of a systematic approach to encyclopedic reference publishing that would define this era. While people had consulted reference books before this, the Britannica became the iconic example of how “look it up” would work for the next two centuries.
When you told someone to “look it up” in this era, you were telling them to navigate an organizational system. Information lived in specific places, and finding it required understanding how those places were arranged. For example, the Britannica organized knowledge alphabetically. If you wanted to know about tigers, you opened the “T” volume and found it exactly where it belonged. The Dewey Decimal System, developed in the 1870s, organized library books by subject. Science lived in the 500s, history in the 900s. A child researching tigers walked to the 599.75 shelf, pulled out a book, and paged through chapters to find what tigers eat.
This navigation process shaped what you learned. Because information was bound in physical pages, you encountered more than what you specifically sought. The child looking for tiger diet information paged past sections on taxonomy, habitat, and conservation. People sometimes learned things they hadn’t set out to find simply because those things lived on the same pages as their target information.
Reference librarians understood this landscape and added a human element to the search process. When someone approached the desk with a question, skilled librarians didn’t immediately point to a shelf. They conducted what the profession calls a reference interview, asking questions like “Tell me what you’re working on” or “What kind of information would be most helpful for your project?” This conversation often revealed that the patron needed something different from what they initially requested. The librarian helped them formulate better questions and navigate the organizational systems more effectively.
However, the barriers to this system were substantial. Encyclopedia sets cost hundreds or thousands of dollars in today’s money, placing them out of reach for many families and smaller libraries. Books became outdated the moment they went to print. And if you needed information after the library closed, you waited until morning. Geographic and economic factors determined who could access comprehensive reference collections and who couldn’t.
These limitations created pressure for change. The cost and physical constraints of print references meant that most people had limited access to information. Libraries could only hold so many books. Encyclopedia publishers could only update their volumes every few years. The requirement to physically travel to where books lived meant that information access depended heavily on where you happened to live. As the World Wide Web became publicly available in 1991 and browsers made it accessible to non-technical users, it offered solutions to many of these constraints.
The Search Era (1998-2020s): When Looking It Up Meant Choosing Keywords
When you told someone to “look it up” after Google launched in 1998, you were telling them to formulate a search query. The fundamental question changed from “Where is this information classified?” to “What words might appear on a page about this?” This shift required a different cognitive skill than navigating physical organizational systems. You no longer needed to know where something lived in a classification system; you just needed to guess what words might appear on a page that discussed what you wanted to know.
People learned to strip questions down to essential terms. “What do tigers eat” became “tiger diet.” This keyword thinking shaped both what you found and how you found it. When you searched “tiger diet,” Google returned pages specifically about that topic. Many people clicked the first link, scanned for the answer, and closed the tab. The process could take thirty seconds. The information about tiger conservation and habitat that might have appeared on the same pages in an encyclopedia didn’t appear unless you specifically searched for those terms. Now, you found precisely what you looked for, which meant you often missed the adjacent information that physical proximity used to provide. The adjacent knowledge that used to come bundled with your answer got unbundled.
Online search engines represented a significant democratization of access for many people, though the expansion was uneven. The kid researching tigers could now find hundreds of sources in seconds from a home computer, assuming reliable internet access. Students in areas with limited library resources could search for information that might not be available locally. The constraints of library hours and physical distance became less limiting for people with home internet connections. A much wider range of information became accessible to many people, though academic databases, specialized resources, and quality internet connections remained unevenly distributed.
The verification landscape also transformed. The Britannica had carried institutional authority. Entries were written by credentialed experts and fact-checked by editors. When you read that tigers are solitary hunters, you trusted that statement because the encyclopedia’s reputation depended on accuracy. Google’s index, by contrast, includes personal blogs, corporate marketing, and peer-reviewed journals with equal algorithmic weight. A search for almost any topic returns thousands of results of wildly varying quality. The internet has no equivalent gatekeeping.
The burden of verification fell entirely to users. Some developed effective strategies: checking whether a site ended in .edu or .org, verifying publication dates, cross-referencing claims across multiple sources. Others struggled to evaluate source credibility and encountered misinformation without recognizing it. The skill of source evaluation became as essential as the skill of search formulation, yet most people developed these abilities informally, through trial and error rather than systematic instruction. People found themselves with unprecedented access to information and unprecedented vulnerability to misinformation simultaneously.
This tension between access and verification, along with the growing volume of information available online, created demand for a different kind of tool. By the late 2010s, people were navigating massive amounts of search results. Finding information had become easy. Finding reliable information required significant skill and effort. What came next would eliminate the search results page entirely.
The Synthesis Era (2022-Present): When Looking It Up Means Asking a Question
When you tell someone to “look it up” after ChatGPT’s public release in November 2022, you’re increasingly telling them to ask an AI. Reports indicate the system reached one million users within days and tens of millions within months, reflecting how dramatically the interaction had simplified.
Large language models represent a fundamental shift in how “looking it up” works. When you ask ChatGPT or Claude a question, you receive something that looks like an answer written specifically for you. The AI doesn’t point you toward sources; it synthesizes information from its training and presents it as coherent prose. The experience feels less like research and more like conversation.
When you ask “what do tigers eat,” you receive prose explaining that tigers are carnivores that primarily hunt large ungulates like deer, wild pigs, and water buffalo. The response appears in seconds, and you can ask follow-up questions in natural language: “How often do they need to hunt?” or “What happens when prey is scarce?” You can say “explain it like I’m five” or “give me more detail about that second point” and the system adjusts. The interface meets you where you are in a way that neither encyclopedias nor search engines could.
For many questions, this proves remarkably efficient. Want to know the difference between baking powder and baking soda? An AI can explain it clearly in seconds. Need help understanding a complex concept? The AI can break it down, provide examples, adjust its explanation based on your background. This accessibility has meaningful implications, particularly for people who struggle with traditional forms of information literacy or who have disabilities that make conventional research difficult.
This conversational quality marks a fundamental shift in what “looking it up” means. Three distinct steps that used to be visible and separate merged into one seamless action. In the encyclopedia era, the source was visible. You held a physical book with the publisher’s name on the spine. In the search engine era, multiple sources appeared, and you chose which to trust based on URL, publication date, and site reputation. You practiced evaluation with every search. In the AI era, the system synthesizes information from its training data and presents that synthesis as an answer. The sourcing becomes invisible. You receive information without seeing where it came from, how it was verified, or what alternative perspectives might exist. Notably, you don’t know what got left out or why.
This invisibility changes the verification process in ways researchers are only beginning to understand. Some early studies suggest that people may verify AI-generated answers less frequently than search engine results. The extra friction of search results, which requires choosing and clicking through to sources, appears to prompt more verification attempts than the seamless answer from an AI. The ease that makes AI powerful for getting quick answers may simultaneously reduce the critical engagement that helps people evaluate information quality.
The reference interview also disappeared from the process. When you asked a reference librarian about tigers, they asked clarifying questions that helped you articulate what you actually needed to know, which often differed from what you initially requested. This back-and-forth helped you formulate better questions and sometimes revealed that you were asking the wrong question entirely. AI systems respond to the question you asked, though modern systems sometimes attempt to identify and address false premises in questions. The quality of this clarification varies. When the system recognizes a misconception, it may correct it. When it doesn’t, it may provide an answer that reinforces misunderstanding.
What Shifts in Each Evolution
Each evolution has traded certain capabilities for others, changing what we encounter when we look something up and how we understand truth and authority.
Serendipity gives way to specificity then synthesis.
In the shelf era, “look it up” sometimes meant encountering information you weren’t specifically seeking. A student researching the Civil War might flip past entries on civil rights on the way to finding battle information. Those adjacent entries could create unexpected connections between topics, spark interests, and reveal relationships between different areas of knowledge.
In the search era, “look it up” increasingly meant finding exactly what you searched for. When you search “Civil War battles,” Google returns pages about battles. You click the first link, scan for the answer, find it, and leave. You won’t see civil rights history unless you already know to search for that connection. You learned precisely what you looked for, and often nothing more.
In the synthesis era, “look it up” means receiving a synthesized answer that presents information with apparent completeness. When you ask an AI about Civil War battles, you get a coherent response, but you don’t see the range of sources that informed that response or the interpretive choices the system made in constructing its answer.
Evaluation becomes invisible.
The shift to AI compressed the evaluation step in specific ways. When you look at search results, you can make choices about which sources to trust based on URL, publication date, and site reputation. You develop critical thinking skills through that evaluation process. When an AI provides an answer, you evaluate the answer itself without visibility into the underlying sources. The practice of assessing information quality based on source characteristics becomes less frequent.
Uncertainty becomes smoothed over.
Transparency about uncertainty also shifts across eras. An encyclopedia might note that “scholars disagree on the exact date of this event” or include a section labeled “disputed theories.” A search engine shows you when very few sources exist for something, or when sources contradict each other. An AI presents information with consistent confidence in its phrasing, even when that information might be based on limited or contested sources. The texture of knowledge, which includes knowing what we don’t know and what remains debated, can get smoothed into seemingly certain statements.
Authority migrates from institutions to algorithms.
As the source becomes increasingly invisible across these eras, the nature of authority transforms.
The encyclopedia era embedded a model where truth was institutional and curated. Knowledge was what credentialed experts agreed belonged in the encyclopedia. This excluded marginalized perspectives, moved slowly, and concentrated power in the hands of publishers and editorial boards. But it created shared reference points. People could disagree about interpretations while agreeing on basic facts because everyone was working from the same source material. When you looked something up in an encyclopedia, you understood you were consulting a compiled reference work.
The search engine era distributed authority across thousands of sources, many contradicting each other. Voices that would never have been published by traditional institutions could reach audiences. Niche perspectives could find audiences. Suppressed voices could be heard. But conspiracy theories and misinformation could also spread alongside fact-based reporting. The burden of sorting through competing claims fell to individual users, most of whom lacked training in source evaluation.
The AI era introduces algorithmic authority. When an AI provides an answer, users often trust it not because they have verified the information, but because the system seems intelligent and confident. People are placing trust in systems whose decision-making processes are largely opaque and whose training data cannot be fully audited by users.
The difference matters because algorithmic decisions are not inspectable in the same way human decisions are. You can read an encyclopedia’s editorial standards. You can click through search results to see what sources exist. But you have much more limited ability to understand how an AI arrived at its answer or what sources it drew from.
When you ask an AI about climate change policy or historical events where interpretation matters, the system makes choices about how to frame the issue. An encyclopedia marked controversial topics as controversial. A search engine shows you the range of positions that exist. An AI synthesizes and presents, which can flatten genuine disagreements into seemingly neutral statements. Those choices are not visible to you. You receive the synthesis without seeing what got emphasized, what got minimized, or what got left out entirely.
People also appear to be developing relationships with AI that can feel like relationships with authority figures. People thank AI systems, apologize to them, and develop trust in them. This shift matters because it affects our critical stance toward the information we receive. When you ask an AI, it can feel like you are consulting an expert who knows the answer, even though the system is producing a statistical synthesis rather than drawing on verified expertise.
Capabilities expand unevenly.
Each evolution has also expanded capabilities in specific ways. Search engines made certain types of information more accessible to people without comprehensive reference collections, though disparities in internet access remained. AI expands accessibility further through conversational interfaces that can help people who struggle with reading complex text, translation capabilities for those who don’t speak the dominant language of internet content, and direct answers for people with cognitive disabilities that make traditional research challenging. The technology adapts to the user in ways that print and search never could.
The speed of access has also compressed timeframes for certain types of learning and problem-solving. A medical researcher can potentially synthesize findings from multiple papers more quickly with AI assistance. A student struggling with calculus at midnight can get immediate help instead of waiting for office hours. A small business owner can get quick explanations of complex regulations. This velocity enables approaches that previously were constrained by time, though the quality and accuracy of that assistance varies.
How Skills Develop Differently in Each Era
The visible process of learning research skills changes across these eras in ways that have lasting implications for how people think and learn. The skills you develop when you “look something up” depend entirely on what that phrase requires you to do.
The Shelf Era Built Organizational and Spatial Thinking
Using a card catalog meant practicing organizational thinking every time you searched for information. You had to understand classification systems: how the Dewey Decimal System grouped related subjects, how alphabetical ordering worked, where cross-references might lead you. People who regularly used card catalogs often developed mental models of how knowledge is organized and categorized. They trained their brains to think systematically about where information might live. Those neural pathways developed through repeated use.
This era also encouraged spatial memory. You might remember not just what you learned, but where you learned it. The information about tiger conservation lived in the middle section of the book with the blue cover on the third shelf. This physical anchoring of information created memory associations that purely digital information doesn’t provide.
Working with reference librarians taught question-formulation skills. You practiced articulating what you knew and what you didn’t know. You learned to describe the kind of information that would be useful rather than asking for a specific source. These metacognitive skills (thinking about your own thinking) could transfer beyond library research to other problem-solving situations.
The Search Era Built Keyword Analysis and Source Evaluation
Search engines required different cognitive work. Instead of navigating a classification system, you had to analyze the language of your question. What are the essential terms? What synonyms might appear in relevant content? How specific should your search be? People who frequently used search engines often developed intuitions about language and associations. They got practiced at guessing which words would appear in relevant content. They learned to scan quickly through multiple sources, extracting key information while filtering out noise.
This era also created opportunities to develop source evaluation skills, though people developed them unevenly. When every search returned dozens or hundreds of results of varying quality, you had to make judgment calls. Does this site seem credible? Is this information recent enough? Does this source have an obvious bias? People who developed these skills became more sophisticated consumers of information. They learned to trace claims to their sources, to identify when content lacked citations, to recognize when different sources were all citing the same original report.
The challenge is that these skills weren’t taught systematically. Some people developed sophisticated evaluation strategies through trial and error. Others never moved beyond trusting whatever appeared first in search results. This created a growing gap in information literacy that existing educational systems struggled to address.
The Synthesis Era May Build Conversational Reasoning
Research on how AI use shapes cognitive development is still emerging, given how recently the technology became widely available. Early observations suggest that frequent AI users may be developing conversational reasoning and iterative question-refinement skills. They practice iterative question-refinement, learning to build on partial answers and adjust their queries based on what the system provides.
If these patterns hold, AI users might develop stronger skills in dialogue-based problem-solving than previous generations. The ability to have a back-and-forth conversation about a topic, to request clarification, to ask the system to explain its reasoning, could translate into stronger collaborative and communication skills more generally.
The concerning possibility is what might not develop. If AI reduces how often people formulate search queries, they may not develop the language analysis skills that the search era built. Without regular practice evaluating competing sources, critical thinking about information quality may not develop as fully. And when answers arrive without requiring navigation of organizational systems, people have fewer opportunities to develop systematic thinking about how knowledge is structured and categorized.
The consequences extend beyond information literacy. The ability to evaluate sources, to recognize bias, to understand how claims are supported by evidence, to identify what remains uncertain or contested—these skills matter for civic participation, for professional work, for making informed personal decisions, and more. If a generation grows up without developing these capabilities because AI removed the need to practice them, they may struggle to distinguish truth from falsehood when it matters most.
What Information Literacy Means Now
We’re not passive recipients of technological change. We make choices about how we adopt and adapt to new tools. Those choices will shape what the AI era means for how humans relate to knowledge.
We can choose to use AI while maintaining practices that previous eras taught us. We can still ask clarifying questions before searching for answers, conducting our own version of the reference interview with ourselves. We can still evaluate the information we receive critically rather than accepting it at face value. We can still seek out multiple perspectives on complex topics. We can still tolerate the friction of not knowing immediately and use that space to think.
The skills that matter remain consistent across eras, even as the tools change. Source evaluation, research methodology, and critical thinking about information quality don’t become obsolete because the interface has changed. Understanding what AI does well, where it has limitations, and how to verify AI-generated information builds on the same foundation as evaluating encyclopedia entries or search results. Information literacy education needs to evolve to include AI interaction strategies while maintaining instruction in source evaluation, research methodology, and critical thinking.
Information inequality appears to be shifting rather than disappearing, too. In previous eras, information inequality centered on physical access to books and libraries, then on internet access and quality. Going forward, information inequality may increasingly involve how people use AI tools. Some people will likely learn to use these tools critically and effectively, verifying important claims and asking probing questions. Others may use them passively, accepting answers without verification or critical engagement. The gap between these approaches could create new forms of disparity.
The instruction to “look it up” is not going away. What we mean when we say it continues to change. Understanding that evolution, with both its gains and its shifts, positions us to navigate what comes next more thoughtfully. We cannot return to the encyclopedia era, and most of us wouldn’t want to. But we can carry forward valuable practices from each era while building the skills we need for the era we are entering. We need to understand what we might be trading so we can decide what is worth preserving. That awareness is itself a form of literacy, one that will matter in the years ahead.
When I was growing up in Chicago, you could actually call the Chicago Public Library reference librarian desk and ask them to research any question for you. You would either hold or they would call you back. I remember using it a few times a year. You could solve an argument among friends. Get help with homework. It was actually quite amazing.
This is a very interesting post! A couple of follow-up thoughts from an AI resister:
- I use a variety of search engines and am finding that the quality of the results has diminished. I am guessing companies are not investing in search because they want to give us their pre-digested AI slop and make us reliant on their version of reality. I turn off "AI summaries" because I am not interested in interested in surveillance-oriented, extractive, exploitive tools.
- I was surprised that you did not include library or journal searches! When I have a question I go to library databases I can access to as a Cornell University alum. If I am looking for open-access sources to share in my newsletters I go to the publisher's journal page, special collections and open, public libraries. The US Library of Congress has digitized a lot of fascinating materials including audio and image files.
Find lots of options here: "Find open-access scholarly journals in the social sciences" https://janetsalmons.substack.com/p/find-open-access-scholarly-journals
Also see: Archival Methods for Online Researchers https://janetsalmons.substack.com/p/archival-methods-for-online-researchers