Just in the span of a few years, the way that we search the internet for information has completely changed. Traditionally, users would conduct a query in a search engine, select links of interest, determine authorship of those pages or articles, inspect domains, and evaluate the evidence across sites. However, in 2023, Google introduced the use of large language models (LLMs) to deliver an AI-generated overview of answers to internet queries–and everything we knew about best practices for searching for information online turned on its heel.
With AI Overviews now in play, we need to change the methods we use to teach K-12 students how to effectively search and interpret sources on the internet. In 2018, Jodi Pilgrim (coauthor), along with Elda E. Martinez, developed the SEARCH strategy, a media literacy framework designed to help students critically evaluate online information. The strategy’s acronym was intended to help learners navigate the internet responsibly and apply critical thinking before accepting digital content as credible. The 2018 version of SEARCH stood for:
- Select key words
- Evaluate
- Add Boolean terms
- Refine results
- Check the URL
- Hunt for important information
Now that most searches include an AI Overview that synthesizes information from multiple sources, however, the SEARCH strategy has lost its effectiveness. Elements like “add Boolean terms” and “check URLs” are obscured or absent, making it difficult for students to apply SEARCH as originally intended. In addition, students may lack insight into how AI responses are constructed and which sources are weighted for importance. Instead of judging credibility across multiple texts, students must now learn to assess the plausibility, bias, and limitations of a generated explanation—a different digital literacy demand altogether.
Digital Literacy Skills, Redefined
Google CEO Sundar Pichai described AI Overviews as, "One of the most positive changes we've done to search in a long, long time.” And certainly there are ways that these overviews are helpful. For example, consider a Google search for "dolphins," which consistently results in over 285,000,000 results. Narrowing a search like this can be a challenging task for children, but with AI powering search engines, a student can ask a question and receive a direct answer written in natural language and organized in a reader-friendly manner.
AI Overviews also change our search behaviors. Instead of searching, we are now conversing. For example, when asking a question about a homework problem, a vacation itinerary, or a DIY project, our query will result in an initial synthesized answer developed from three to five sources, rather than a list of websites. What follows the synthesis is an invitation to "dive deeper in AI mode" for additional information. That invitation initiates a conversation with AI to refine results. Where keywords were once essential to an effective online search, now even vague questions can generate detailed responses, with links to explore for further examination. Detailed queries across one or more tries, or conversations, has become a skill known as "prompt engineering." In addition, we don't have to skim and analyze the results ourselves. We can request a quick summary of an article or a topic, and AI will generate it in seconds.
AI Overviews change our search behaviors. Instead of searching, we are now conversing.
Given this new landscape, students need similar, but updated, digital literacy skills. AI-generated search overviews require students to think critically about a number of factors: Where are these AI outputs coming from? How are the data sets trained? What is the environmental impact of these searches? It also requires students to cross-check AI-generated claims and understand how AI systems gather information. Therefore, AI literacy needs to be a skill our students learn early and often. Even if students are discouraged from using AI platforms in school, they will still encounter AI outputs in their daily internet use–and they need to learn how to evaluate them responsibly.
As educators, we must consider how to teach students to navigate this new way of searching–and how AI Overviews can be problematic. For example, we know that generative AI can produce inaccurate or fabricated content. How do we ensure students understand how AI works and how to critically evaluate the information it provides? How do we teach them to make ecological choices, such as knowing when a simple internet search or checking a book out of the school library would consume less energy and more more effective that an AI search?
With all these questions in mind, we now propose a modified SEARCH strategy to support students in this AI-influenced environment.
A Revised SEARCH Strategy
Search Question
Start a search with a question. A clear, precise search question will yield a more developed response. For example, asking What are the best board games for children ages 7-10? will produce a better answer than a query like, What are the best games? We recommend using a search engine rather than an AI platform, since generative AI platforms use about five times as much electricity as a simple web search. In addition, the clearer your question is, the more energy-efficient your search process will be. You can also add “- AI” to your search to remove the Overview from web results completely. (For more information on sustainable uses of AI, see Eco-Conscious Ways to Use AI.)
Evaluate Initial Result
As an AI Overview is generated, the learner must evaluate it and its accompanying links, just as in the original SEARCH model. Does the AI Overview include links to recent and relevant article sources? Are they linked to credible sources? AI literacy skills now come into play as learners consider how the AI content was produced. In Google, the most widely used search engine, results provide built-in transparency for users. Other search engines are working toward AI summary transparency, or at least giving a warning of the need for users to evaluate initial results, but teachers should always caution students to examine the sources.
To do so, students need to do a bit of detective work. In Figure 1, for example, we started a search query, “What created the Dust Bowl?” This question generated an AI Overview. By clicking on the three dots in the upper-right corner of the results (highlighted below), we can then learn more about the sources and the process used to develop the Overview.
Figure 1: AI Overview Results for Dust Bowl Search
Once we click on the three-dots, we can view detailed information about the source used in the overview. (See fig. 2)
Figure 2: Dust Bowl Detailed Source Information
Students should look for and prioritize websites from respected organizations. In the Dust Bowl example, students should review the links in the AI Overview or from the source “bubbles” or “cards” located on the right sidebar (See fig. 3) and search for reputable organizations such as national associations and education services, museum sites, and state historical societies. Teachers should also show students how to find the author or publisher of a website, and their “about” information. If the information needed is found at this point, there may be no need to engage AI further, and the search can stop and save energy use.
Figure 3: Source "Bubbles" or "Cards" from 'What Created the Dust Bowl?" Search
Ask AI
If the initial search does not provide relevant information, it may be time for “Dive deeper into AI Mode.” Continuing to “dive deeper” into the Dust Bowl, a student asks, “What is meant by "severely affected areas"? This question leads to more information about top soil loss and also new information about the mass migration of people from the affected areas to California and how the dust buried roads, fences, and whole buildings. It includes a map of the affected state areas. Generative AI then offers to continue the conversation by asking additional questions the user may consider and additional research offers, like, “Would you like to explore the migration routes these families took to California or details on the soil conservation techniques used to recover the land?”, with bolded phrases emphasizing important topics within the overarching event. The student can choose one or both of these questions to “dive deeper” into the content.
Read Results
After students evaluate the initial websites used to provide the AI Overview or Deep Dive, they must ensure the search yielded a relevant response. This step prompts students to pause and make sure they read and understand the answer to their question. In other words, what is the main idea of the response? AI Overviews simplify the search process by producing answers in short narrative paragraphs and bulleted concepts. Key ideas are typically listed, but teacher modeling of how to find the key ideas may be necessary. In addition, for children with barriers to reading text, a teacher may show them how to listen to it using text-to-speech applications. As the response is read, students should consider whether or not the response seems true. Teachers should explain that AI sounds confident, even when it is wrong.
AI-generated search overviews require students to think critically about a number of factors: Where are these AI outputs coming from? What is the environmental impact of these searches?
Cross-Reference
As students consider the accuracy of the response, the next step in the search process is to check to see if multiple sources say the same thing. This is called cross-referencing, and it is important because the websites used in the AI Overview may be missing information. A teacher should prompt students to check important facts to see if they can find the same results somewhere else. For instance, after reading the Dust Bowl response, a follow-up search may be, “Did the Dust Bowl strip the Great Plains of native grasses?” Checking for consistency of information across sources verifies facts and identifies missing information.
Hunt for Bias or Hallucinations
As students read and cross-check their search results, they should check the information for potential bias—one-sided or misleading content—and/or AI hallucinations—misleading or inaccurate output generated by AI tools. This requires students to think critically about AI and its potential to deliver fake facts. Detailed specifics provided without evidence constitute a red flag. For instance, if an AI overview states that “97% of dolphins live in fresh water,” a student should immediately wonder where the number came from and if another source can verify the statistic. In addition, when AI provides titles or dates, they should always be fact-checked.
The Future of SEARCHing
Internet searches will continue to transform. They will probably get faster, have fewer hallucinations, and hopefully become more energy efficient. But even as the machines get better, humans must remain in the process. Our updated SEARCH strategy provides a new and revised response to generative AI’s influence and a way to ensure that students think critically about the information they seek and are given. Using this strategy can help educators foster the development of responsible digital citizens.
