Google Discover & Top Stories: How They Work

When Search Comes to You

Most people think of search as something they initiate. A question forms, fingers type, results appear. But two of Google's most widely used surfaces work differently. Google Discover and the Top Stories carousel don't wait for a query. They surface content before anyone asks for it, based on what search engines already understand about a person's interests, behaviors, and the entities that matter to them. Understanding why this is possible, and how it works, changes the way you think about what search has become.

The Query-Free Surface: What Google Discover Actually Is

Google Discover is a content feed that appears on the Google app homepage and on mobile browsers when users open a new tab. It presents articles, videos, and other content without the user ever typing a word. To anyone who grew up thinking of Google as a search box, this feels counterintuitive. Search without searching.

The mechanism behind it is an interest graph built from everything Google knows about a person. Search history, watch history on YouTube, location signals, app usage, the topics of articles previously read, and the entities a person has shown consistent interest in all feed into a model of that person's ongoing curiosity. Discover then matches that model against a continuously updated index of fresh content.

This is not recommendation in the Netflix sense, where the system tries to predict what you'll enjoy based on what similar users liked. Discover is more directly tethered to demonstrated interest in specific topics and entities. If someone consistently reads about electric vehicles, follows Formula 1 results, and searches for climate policy, Discover builds a picture of those interests as named concepts, not just vague preferences, and surfaces content that connects to them.

Entities as the Bridge Between Person and Content

The reason Discover can work without a query is that Google's understanding of both people and content is built around entities. An entity is a named thing in the world: a person, a place, an organization, a concept, an event. Google's Knowledge Graph contains billions of these entities and the relationships between them.

When Google indexes a piece of content, it doesn't just record keywords. It identifies which entities the content is about, how authoritatively it covers them, and how those entities connect to related concepts. When Google builds a user's interest profile, it similarly maps demonstrated interest onto entities. The bridge between person and content is therefore entity alignment: does this content relate to the entities this person has shown interest in?

This is why entity-based content understanding matters so much in modern search. Content that clearly signals which entities it covers, and does so with depth and authority, is more legible to a system trying to match it to the right audience proactively. Vague or keyword-stuffed content is harder to place in an entity framework, which makes it harder to surface in query-free environments like Discover.

Freshness, Engagement, and the Discover Ranking Model

Discover doesn't rank content the way traditional search results do. There is no query to match. Instead, the system weighs several signals to decide whether a piece of content is worth surfacing to a specific user at a specific moment.

Freshness matters more in Discover than in most search contexts. Because the feed is about what's happening now in areas of interest, recently published content has a structural advantage. But freshness alone is not enough. A fresh piece of content about a topic the user has never shown interest in won't appear. The interest signal must exist first.

Engagement signals also play a significant role. Content that earns strong click-through rates, long dwell times, and low return rates (where users don't immediately bounce back to the feed) signals that it genuinely satisfied the interest it promised to address. Over time, this shapes which publishers and content types Discover favors for which interest profiles. The system is essentially learning what good looks like for each person and each topic cluster.

Page experience signals matter here too. Discover has historically penalised content that loads slowly, uses intrusive interstitials, or delivers a poor mobile experience. This isn't arbitrary. A feed that consistently delivers frustrating experiences loses the trust of the person using it, so the system has a structural incentive to favor content that performs well technically.

Top Stories and the News Ecosystem

The Top Stories carousel operates differently from Discover but shares the same foundational logic: surfacing content based on what's happening now in relation to what a user cares about or is currently searching for.

Top Stories appears within search results, typically near the top of the page, for queries that have a news dimension. A search for a politician's name, a breaking event, or a topic with active public discourse will often trigger the carousel. Unlike Discover, Top Stories is query-triggered, but it selects from a curated pool of content that Google has determined meets its news quality standards.

Inclusion in Top Stories depends on a combination of factors. The content must be indexed quickly, which means Google's crawlers need to be able to access and process it without delay. The publication must have established signals of authority and trustworthiness in its topic area. The content itself must be original reporting or substantive commentary, not a thin rewrite of something published elsewhere.

Google News, which feeds into Top Stories, applies what it calls a prominence and authority model: publications that consistently produce original, accurate, widely cited journalism earn greater prominence over time. This is an entity-level judgement about the publication as an organization, not just a page-level quality assessment.

Why These Surfaces Represent a Shift in How Search Works

Discover and Top Stories are not peripheral features. Discover alone reaches hundreds of millions of users monthly. For many publishers, it drives more traffic than traditional search. Understanding why these surfaces exist and how they function reveals something important about the direction search has taken.

Traditional search is reactive. A need arises, a query is typed, results are returned. Discover is anticipatory. The system tries to surface content before the user knows they want it, based on patterns of interest that persist over time. This shift from reactive to anticipatory search is only possible because Google's understanding of both content and people has become sophisticated enough to operate at the entity level rather than the keyword level.

Top Stories represents a different kind of evolution: the integration of real-time information into a system that was originally built around static documents. The web changes constantly, and news is the most time-sensitive expression of that change. Google's ability to surface relevant, authoritative news content in near real-time reflects how deeply the crawl, index, and ranking systems have been engineered around freshness and trust.

What This Means for Understanding Search as a System

The existence of Discover and Top Stories challenges the assumption that search is fundamentally about matching queries to documents. These surfaces show that search has evolved into something broader: a system for connecting people with information that is relevant to their interests, needs, and current context, whether or not they have articulated those needs as a query.

This has implications for how content authority is understood. A publisher that has built genuine expertise in a domain, produced original work, earned citations from other authoritative sources, and demonstrated a consistent track record of accuracy is positioned to benefit from both surfaces. These are not signals that can be manufactured quickly. They accumulate over time as a function of real editorial quality and genuine audience relevance.

Understanding Discover and Top Stories also clarifies why entity clarity in content matters so much. A system that matches people to content based on interest graphs and entity relationships needs content to be legible at the entity level. Content that is clearly about something, that names and contextualises the entities it covers, that connects those entities to related concepts with accuracy and depth, is far more useful to a proactive surfacing system than content that is optimized around keyword density without semantic coherence.

Search, in this light, is not a box waiting to be queried. It is an ongoing model of the world and of the people in it, continuously updated, continuously matching. Discover and Top Stories are where that model becomes most visible.