How AI Is Changing Search Behaviors

Source: Nielsen Norman Group · 2025-08-15

Generative AI (genAI) is reshaping how people search for information. Anyone watching their content pageviews decline is currently experiencing the impact of this. But what’s behind the shift? The speed of the change is impressive, considering how deeply ingrained information-seeking habits can be.

In a recent qualitative study, we asked people to bring their own research tasks into the virtual lab. We explored how users' information-seeking behaviors are shifting in response to AI-powered search tools and chatbots. While AI offers compelling shortcuts around tedious research tasks, it isn’t close to completely replacing traditional search. But, even when people are using traditional search, the AI-generated overview that now tops almost all search-results pages steals a significant amount of attention and often shortcuts the need to visit the actual pages.

Information-seeking habits are sticky. Once someone finds a reliable way to easily find information they need, that method becomes nearly instinctive.

This partly explains Google’s massive share of the search-engine market. Over the years, many participants have told us they’ve never even considered Bing or other alternatives, simply because they’re familiar with Google and know it works well (enough) for them. In our study, multiple participants commented on their tendency to lean on what they’re already comfortable with.

People tend to reach for whatever information-seeking tool is most convenient. One participant reflected that he got into the habit of using Google because it was built into his Chrome browser.

Users rely on them because they’ve generally worked well in the past. To change these habits, people need a significant incentive.

Generative AI’s value in information seeking is powerful enough to change those ingrained habits. Several participants were aware that their own information-seeking behaviors had started to shift since they began experimenting with AI tools.

“Oh, I always start from Google. I never use different search engines, but these days I also incorporate ChatGPT.”

Generative AI offers substantial shortcuts around the often tedious and time-consuming work required to research a topic, including:

Even when participants utilized only a handful of genAI’s possible information-seeking benefits, they valued the assistance immensely.

AI overviews appear at the top of results pages for many queries on web search engines. They’re powered by LLMs and attempt to quickly define keywords or answer questions.

AI overviews are the modern, upgraded iteration of the featured snippets and answer boxes introduced by Google and its competitors in the 2010s. Even then, many content sites started to notice a dent in their web traffic as people started to find answers without a click. AI overviews are even more likely to satisfy information needs without clicks.

(This feature presents a serious challenge for all content producers — NN/G included.)

A recent quantitative study by Pew Research analyzed this effect — they found that Google searchers who encountered an AI overview were substantially less likely to click on results links.

But AI overviews are limited in their usefulness — they’re better for quick definitions and fast facts (sometimes incorrect facts) than for complex information-seeking needs that involve synthesis. In these situations, AI chatbots (ChatGPT, Gemini, Google AI Mode, Perplexity, etc.) were more useful. But that use case for AI isn’t obvious to everyone.

Our study included a few participants who experienced AI chat’s info-seeking benefits for the first time during our study sessions. We watched as their information-seeking habits shifted in front of us.

For example, one participant had read AI overviews on Google and frequently used ChatGPT to write and edit emails for work but had never considered using it for information seeking until our study.

When shopping for a football (soccer to the Americans) goal for his 11-year-old son, he used Google search to find the dimensions. An AI overview helpfully supplied them.

The participant found the AI overview useful, but his utilization of AI ended there. He proceeded with his task the “old-fashioned” way — manually searching, visiting websites, and scanning web pages. When considering each option, he wanted a recognizable brand and lightweight-yet-durable construction in the correct size and in his price range. He read customer reviews and flipped through product photos. Whenever he found an option he liked, he wrote it down on a physical sticky note. The process took him 10 minutes and would probably have gone on longer if our facilitator hadn’t ended the task early for followup questions.

For a later task (diagnosing a plumbing problem in his house), we directed the participant to Gemini. It was his first encounter with Gemini, and his first time using an AI chat for information seeking. He struggled at first to communicate his precise problem but eventually was happy with the guidance it provided.

“It feels like it saved me a bit of time. It’s drawn in a lot of data and kind of tailored it to what my specific need is. [...] There’s a lot of information out there. Plumbing is a common problem. So, there’s a lot of information to sift through to get to the very specific problem I have. I think it’s done a good job.”

“I'll definitely use this in [the] future. I realize maybe I should have come to Gemini looking for the goals. In fact, I might do this when I get off the line and see what it says about the best places to buy regulation goals in the UK.”

While generative AI does offer enough value to change user behaviors, it has not replaced traditional search entirely. Traditional search and AI chats were often used in tandem to explore the same topic and were sometimes used to fact-check each other.

All our participants engaged in traditional search (using keywords, evaluating results pages, visiting content pages, etc.) multiple times in the study. Nobody relied entirely on genAI’s responses (in chat or in an AI overview) for all their information-seeking needs.

The participants in our study who were already using AI for information seeking reported being most familiar with ChatGPT and Gemini. In the habit-driven market of information seeking, this familiarity is a huge competitive advantage. In this and previous studies on AI usage, some participants were calling ChatGPT simply “Chat,” which is reminiscent of how Google became a verb (“to search” was “to google” regardless of the search engine). Linguistic shifts can portend behavioral shifts.

ChatGPT was the first modern LLM chat on the scene. It captured public attention and currently dominates the AI-chat market. But Gemini’s association and integration with traditional Google search gives it a solid chance to catch up.

These early days will be critical in the AI wars. Companies have a lot to gain from becoming users’ habitual go-to for information seeking.

Generative AI tools are undeniably useful for information seeking, and even novice users recognize that potential value right away. However, this study reminds us that, while people working in tech may feel that genAI’s uses are obvious, that is not the case for many consumers.

Discoverability remains a major challenge for genAI design — not only the discoverability of the tools, but also the discoverability of its possible functions and how to use it.