What Is Information Seeking?

People want answers. They might be looking for a review of a local business, a product to purchase, reliable information about a specific topic, and so much more. Whatever the question, for the last thirty years, search engines and websites have been the predominant means of getting an answer.

Generative AI changes this approach to answering questions. ChatGPT, Gemini, Perplexity and Claude can also help you find the information you need to answer your questions. In doing so, they offer a new opportunity for businesses, but these tools also compete with our website for traffic. For many websites, Google’s AI Overviews have already made that traffic concern feel more real.

To understand why generative AI competes with websites requires understanding how people seek out answers to questions. There are various studies about information seeking: how do people ask questions, how do people find the information related to those questions, and how do people bring that information together to form an answer?

Information-seeking models are the best way I’ve found to understand how generative AI changes SEO. The more we understand these models, the easier it becomes to explain why generative AI feels useful, why websites may lose traffic, and where websites still have a role.

In this article, I’ll provide a brief overview of a few different information-seeking strategies and what these suggest for the future of SEO and AI search.

The Three Core Steps: Find, Evaluate, and Aggregate

One of the simplest ways to understand information seeking is through three fundamental steps outlined by the Nielsen Norman Group: finding, evaluating, and aggregating.

1. Find: The goal of this step is to locate potential sources of information that might answer a question. Search engines primarily focus on returning a highly relevant list of links, simplifying the process by filtering out irrelevant data. Generative AI can serve a similar purpose, returning a list of sources like a search engine (as seen in Google’s AI Overviews or Perplexity), or it can allow users to skip this step entirely by providing a direct answer without a list of sources.
   
   
   
2. Evaluate: Once a list of sources is found, users must evaluate them to see if they hold the desired answers. Search engines help by ranking relevant results, but the user still needs to click and verify. Generative AI differentiates itself here because it provides a fully formed response drawn from various sources, including training data or data retrieved during response generation. The upside is the AI tool can greatly simplify the user’s workload because users do not have to click through to the website. That said, by simplifying the user’s workload, that means users are less likely to click to our website (reducing our web traffic). From what I’m seeing working with clients, users may still choose to verify certain types of sources, especially in high-stakes situations.
3. Aggregate: The final step is synthesizing the collected information into a cohesive answer. With traditional search engines, the user must open multiple websites and piece the information together themselves. Generative AI, however, does this aggregation work for the user, offering what is known as targeted aggregation. Unlike a Wikipedia page that provides generalized aggregation about an entire subject, generative AI provides a personalized, specific response across millions of sources instantly.
   

This is why generative AI changes how we need to approach SEO. If users can skip parts of the finding and evaluation process, there may be fewer clicks to websites. But it also means companies need to understand when users still want to come to a website. People may want to come to a website because they want to verify an answer, when they want more details, or when they may want to complete an action (like making a purchase). 

Personally, I don’t see this ending websites as we know them. Instead, this seems more like it is shifting how websites will be used and where websites fit in the information-seeking process. What we are helping clients with at Elementive is understanding exactly how this is changing with that client’s industry and testing out what changes need to be made to support new user behaviors.

Kuhlthau’s Information Search Process

In the early 1990s, Carol Kuhlthau studied the Information Search Process, analyzing a user’s feelings, thoughts, and actions while seeking information. There are six stages to this process:

• Initiation: The whole process starts when you ask a vague question or state an approximate need for something. There is a lot of uncertainty.
• Selection: The next step is to narrow down the question. Kulthau talks about how optimism increases through relevant exploration.
• Exploration: This is the stage where the person searching reads, processes, and evaluates sources. This can be difficult and lead to confusion.
• Formulation: The search becomes more focused, moving into more detailed prompts and longer queries. There is more clarity and less confusion.
• Collection: Next, you start finding specific information that seems like a better answer to the question (i.e. these websites help more than others).
• Presentation (or Action): Finally, you have the information you need or know how to take action.

There are a lot of steps, but we have all experienced this. You start with a question, start trying to search for something related to that question, get frustrated by too many options or a lot of irrelevant options, eventually narrow in on what you actually need to know, and then can make a purchase or get an answer to the question.

All six steps are highly visible when using a search engine like Google, as the user must manually select, explore, and formulate a perspective. In contrast, generative AI tools handle most of the selection and exploration steps on behalf of the user behind the scenes. If anything, Kuhlthau’s framework highlights one of the main benefits of talking with AI about a question instead of doing the search yourself. We all want to move toward clarity and focus to get our questions answered as quickly as possible. Generative AI can do that by taking us directly from the Initiation stage to the Presentation stage.

Evolving Search and Berrypicking

Traditional models often portray search as a clean, linear journey, but Marcia J. Bates described an alternative “evolving search” in 1989 to reflect real-life, non-academic scenarios. Evolving search is iterative and non-linear, where users adjust their queries and search paths based on what they discover along the way. Because information is collected in pieces from multiple sources, Bates compared this approach to “berrypicking”.

You might start by searching for ‘best accounting software for small business’ and open a few of the websites that rank. You can then hop over to Reddit to read reviews before watching relevant YouTube videos. Then you ask ChatGPT how one tool compares to another. Finally, you visit vendor pricing pages. That is berrypicking.

Most traditional Google searches are a form of berrypicking. We are searching Google to find multiple sources we can pick information from. Responses from generative AI tools facilitate this as well. Users will review an AI response, which prompts a new, completely different question to explore more aspects of a subject. The berrypicking concept suggests that rather than strictly competing, it is possible that users will combine conversations, search engine results, and websites together.

Information Foraging

To understand why people choose specific strategies while seeking information, Peter Pirolli and Stuart Card introduced information foraging theory in 1999. Information foraging takes an evolutionary perspective, comparing the search process to an animal foraging for food. When searching, users attempt to maximize their gains of valuable information while reducing the costs (like time and attention) of acquiring it.

People move between “information patches,” which could be a Google search result, a conversation with ChatGPT, or a specific article. Users constantly balance costs and benefits, deciding whether to move to a new information patch or engage in “enrichment” (refining an existing search or using a follow-up prompt in an AI tool). Ultimately, information foraging dictates that users will prefer whichever tool delivers valuable information at the lowest cost. ChatGPT might present the information at a lower cost than a traditional Google search result or than your website.

Key Takeaways

There is a fundamental truth that is hard for me to accept as a technical SEO consultant who optimizes websites: people do not inherently want to use a search engine, visit a website, or talk with an AI tool. All of us simply want to get an answer to a question. 

Accepting this truth changes how we see SEO. For a long time, SEO has been about earning rankings and getting people to come to our website. Rankings and traffic, though, aren’t really where we need to focus. Instead, our focus should be on how we make sure our business is part of the answer people are seeking out. Those answers may appear on our website, in a traditional search result, or as part of a mention in an AI-generated response. SEO now is about making sure our business is represented in all these places. 

Because generative AI can aggregate information into relevant, complete answers, people may no longer need to spend as much time finding and evaluating multiple websites on their own. That is why some website traffic will be lost. It would be easy to see this as a competition between your website and AI responses. To a certain extent, I agree, but AI responses also create a different kind of opportunity: making sure your content, expertise, and brand are part of the information AI tools use when forming those aggregated answers.

Need Help?

If you need help figuring out how to prepare for generative AI and the way it will change SEO, contact me. We can work together to determine how your users are using generative AI tools and what the implications are for your business.