Do people turn to an AI assistant before searching the web, or only after they've already started browsing?

This explores whether AI assistants fit into a person's web journey differently than search engines do. The question is whether people reach for assistants at a different point in their browsing, and whether assistant use replaces search or runs alongside it. The corpus says they do fit in differently, and in a direction you might not expect. A cross-surface panel study that tracked the same users across search, browsing and assistants found that search tends to open a journey toward content, while assistant sessions more often come after it. Assistant use followed earlier web activity 20.6 percentage points more often than it preceded it Do people use AI assistants before or after searching?. That undercuts the popular 'answer engine' story, where people ask a chatbot first and never visit a website. In practice, many people browse first and then bring what they found to an assistant, perhaps to digest, compare or act on it.

The same study adds a twist: within the same users, sessions that used only an assistant were more common than sessions that used only search Do people use AI assistants before or after searching?. So assistants show up in two modes. Some sessions are self-contained, with the user never leaving the chat. Others are follow-ups attached to web activity. Search mostly works as an entry point. Qualitative work fits this picture. Nielsen Norman Group found that every participant kept using traditional search throughout their tasks, often running chat and search side by side, and the main thing holding back AI use was not knowing when it would help, not distrust Does generative AI chat actually replace traditional search?.

The pattern also runs the other way: assistant-style content is moving into the search page. Eye-tracking research shows AI Overviews now take the visual 'golden triangle' that the top-ranked result used to hold. Attention to the first result fell from 31% to 9%, and trust ratings stayed equally high for both Where do searchers look when AI Overviews appear?. So the line between an 'assistant session' and a 'search session' is blurring from both directions. People carry web content into chats, and search pages put an assistant-style summary in front of the links.

Why might the order matter? An 8-day field experiment found that people who learned with ChatGPT did worse than Google searchers, especially on critical-thinking questions. The authors traced this to less agency over which sources to read and less exploration of the wider topic Does ChatGPT harm informal learning compared to Google Search?. Read alongside the sequencing data, this suggests that the common habit of browsing first and then asking the assistant may protect people. It keeps the exploring step that assistant-only sessions skip. That is an inference: none of these studies tested it directly.

A caveat: the corpus has one study that measures the interleaving directly. The others tell you what people do around it (running tools in tandem, where attention lands, what gets learned), not the fine-grained order of steps within a session. If you want the hard numbers, start with the panel study. Read the others for why the pattern might matter.