Google’s Liz Reid recently sat down for an interview that quietly clarified something a lot of SEOs have been guessing at: why some searches land on AI Mode, some stay on classic results, and some never touch AI at all. Her answer centers on a term most people outside Google have never heard — “browsy queries” — and it changes how you should think about which pages deserve full answers versus which ones should stay built for browsing.
Search behavior isn’t one thing anymore
Host Joe Wazenthal asked Reid directly about how people choose between classic search, AI Mode, and the Gemini app. Her first move was definitional: she grouped classic search and AI Mode together as “Search,” and positioned Gemini as something categorically different.
From there, her core point was that there’s no single dominant behavior pattern. A large share of users move across all three surfaces depending on the task, not out of habit but because the query itself calls for a different tool. Informational queries — the kind SEOs have optimized for forever — tend to land on classic search or AI Mode. Creative or productivity requests, like “rewrite this to sound more formal,” skew toward the Gemini app. And within AI Mode itself, there’s a split worth paying attention to: some people arrive by starting in AI Overviews and clicking through, while others go straight to AI Mode because they expect a longer, more involved exchange with follow-up questions.
What a “browsy query” actually is
Then came the phrase that gives this piece its name. Reid noted that when a query is what she called “very browsy,” people tend to prefer the full search results page over AI Mode. It’s not a term Google has defined publicly in a blog post, but it shows up in enough places to reconstruct a working definition.
A former DeepMind and Google engineer described on LinkedIn having built a machine learning model to identify “browse intention” queries — things like “best places to visit in Orlando” — and using that classification to surface more engaging content directly on the results page, which lifted global click-through rate by 5%. Google also uses the phrase in a commerce engineer job listing, describing browsy queries as part of the complexity that shopping search has to handle alongside hierarchical schema and short multimodal documents. And a Google support page for shoppable video ads places “browsy” placements earlier in the shopping journey, before a user has settled on what they want.
Across all three, the common thread is discovery-stage intent. A browsy query isn’t a request for a specific answer — it’s a request to explore. The user hasn’t narrowed down what they want yet, and they’re looking for options, inspiration, or a range of results to scan rather than a single synthesized response.
Why that distinction matters for optimization
If a query is browsy, the traditional SERP format — a scroll of links, images, and options — genuinely serves the user better than a single AI-generated answer would. That has a direct implication for how you structure content aimed at exploratory intent: broad category pages, “best of” roundups, and visually rich listing pages still have a clear job to do. A page targeting something like “stylish outfits for summer” benefits from a pyramid structure, where the entry point stays broad and specificity increases as the user goes deeper into the page or site. Trying to force a browsy query into a single definitive answer misreads what the searcher actually wants.
Query fan-out and the end of keyword-first thinking
The more consequential part of Reid’s interview dealt with what happens to complex queries once they leave the keyword era behind. For decades, users compressed genuinely complex needs into short keyword phrases because that’s what the interface demanded — “restaurants in New York” instead of “a restaurant for five people, not too pricey, with a vegan option and something kid-friendly nearby.” Google’s ranking systems then had to infer the hidden intent behind that compressed phrase, historically using click behavior to figure out which meaning of an ambiguous term was most popular and ranking accordingly, regardless of how many links a page targeting a less popular interpretation had earned.
Reid described AI search as removing that compression step. Users increasingly type the actual, multi-part need instead of translating it into keyword-ese first, because the system is now expected to do that translation itself. The practical mechanism behind this, in her description, isn’t that Google’s AI just reads one long query and answers it. It breaks the natural-language request into smaller, specific sub-queries — a process sometimes called query fan-out — and fires each of those off to classic search infrastructure. The AI layer then synthesizes an answer by pulling from the top few results for each of those smaller queries.
That mechanism has a real consequence for SEO strategy, and it cuts against the instinct to chase every long-tail phrasing: since fan-out still routes through classic search for each sub-query, ranking well for the underlying specific queries still matters more than trying to match an entire long, one-off natural-language question verbatim. A complex AI Search prompt is often unlikely to repeat in that exact form, which lowers the value of optimizing narrowly for it. What doesn’t lose value is precision on the individual, well-defined sub-topics a complex query decomposes into. This is the same shift we covered when Google discussed keyword fragmentation and how it’s reshaping what “ranking for a query” even means.
There’s also a competitive-visibility angle. Because a single AI Overview slot often draws from multiple sources rather than crowning one winner, brand and format signals now carry weight beyond text relevance — a recognizable brand treatment, well-chosen images, and video that can claim its own space in the answer all help a page compete for a shared slot rather than a single ranked position.
Why this makes Google’s job on quality harder
Reid was candid about the operational cost of this shift inside Google. When most users typed similar short keywords, Google could cache results efficiently because so many queries were effectively identical. As natural-language queries diversify, that caching advantage erodes, and Google has to do more real-time work to decompose each unique question, understand its parts, and manage the latency of assembling an answer from several sources instead of serving a pre-computed result.
That operational strain doesn’t show up in Search Console the way older changes did — the queries and impressions data itself has shifted in ways worth understanding if you’re trying to measure any of this, something we dug into in how Google quietly changed search term reporting for some AI queries.
The actual takeaway for SEOs
Reid’s framing of “addressing the real need” can sound like a vague platitude, but it maps onto a genuinely useful audit question: does this page solve the underlying problem, or does it solve the keyword? A page can be technically clean, well-structured, and still fail because it answers the query the user typed rather than the need they actually had. That question — what specific need does this page fill, and how is it different from and better than the next-best page trying to fill it — is a more durable diagnostic than a keyword checklist, particularly as more of search traffic routes through systems that reward being cited as part of a synthesized answer rather than being clicked as the top blue link. It’s also worth remembering that a meaningful share of queries still resolve to open-web links rather than an AI synthesis, a pattern Google’s own data on how many queries get sent to the open web makes clear — full SERPs and browsy intent are not going away.
Frequently asked questions
What is a “browsy query” in Google’s terminology?
It’s an internal-sounding term for a query with discovery-stage, under-specified intent — the user wants to explore options rather than get a single direct answer. Examples include broad requests like “best places to visit in Orlando” or “stylish outfits for summer.”
Why do browsy queries favor full search results over AI Mode?
Because the user hasn’t narrowed down what they want yet. A scrollable page of varied results, images, and options serves an exploratory need better than one synthesized answer would, according to Liz Reid’s description of user behavior patterns.
What is query fan-out?
It’s the process by which Google’s AI systems break a long, complex natural-language question into smaller, specific sub-queries, run each through classic search, and then synthesize an answer from the top results of each. It means ranking for specific, well-defined queries still matters even in an AI-first search flow.
Does this mean long-tail keyword optimization is dead?
Not exactly. Optimizing narrowly for an entire long, one-off natural-language phrase has diminishing returns because such phrases rarely repeat verbatim. What still pays off is precision on the specific sub-topics that a complex query decomposes into through fan-out.