What Is Google Discover and Why You Can't Optimize It for a Keyword

Google Discover is part of Google Search and shows each user content matched to their interests, based on their Web and App Activity. It appears as a feed inside the Google app, on the Chrome mobile home screen on Android and iOS, and in some cases as a widget on Android home screens. None of that requires a typed query: the feed loads on its own, filled with content Google predicts will interest that person at that moment.

That missing query is the real difference from a classic search result. Someone who searches "lentil soup recipe" hands Google a string of text, which gets compared against millions of pages to rank the best matches. Discover has no such string to compare against: the system relies on interest signals, such as search history, location, or the apps someone uses, to decide what to show. Without a query, there is no keyword to rank a page against either. Framing Discover as "optimizing for keyword X" describes a mechanism that simply does not exist in this channel.

That does not mean any indexed page automatically qualifies. Google requires the same baseline as for normal search: the page has to be indexed and follow the general policies. At the same time, Google states explicitly that meeting those requirements does not guarantee appearing in the feed. Eligibility is a minimum condition, not a promise of visibility.

In practice, that produces an effect that surprises a lot of site owners: two articles on the same topic, both cleanly indexed with no technical errors, can perform completely differently, with nothing in Search Console explaining why. For classic search, that kind of gap can almost always be traced to backlinks, content depth, or search intent. Discover offers no such anchor, because there is no query to measure the difference against. That makes Discover harder to analyze than regular search, not just harder to influence.

The most common confusion around Discover comes from treating it as a variant of regular search. The table below shows where the two overlap, and more importantly, where they don't:

The row most people overlook is the last one: while classic search also exists on desktop, Discover is essentially a mobile product. That shapes any image or format strategy built around it.

Google does not publish a formula for Discover, but it does describe a handful of factors in detail that increase the odds of appearing, always framed as guidance, never as a guarantee. The most concrete one is technical: the featured image must be at least 1200 pixels wide, exceed 300,000 total pixels of resolution, and use a 16:9 aspect ratio, specified through the og:image meta tag or schema.org markup. A generic logo or a decorative image unrelated to the content lowers the odds of that article being shown, even when the writing itself is excellent.

The second block is editorial quality: headlines that describe the content accurately instead of manufacturing false expectations, no clickbait tactics or sensationalized titles, and topics that are either genuinely timely or offer a perspective that stays useful past the news cycle. A headline like "You won't believe what happened next" breaks this rule even when the article itself is solid: it sets up an expectation the text doesn't deliver on, and that is exactly the kind of sensationalized tactic Google's documentation warns against. Google explicitly links its Discover documentation to its guidance on creating helpful, people-first content, the same framework that underpins E-E-A-T: anyone chasing results in Discover should read those criteria, not a list of Discover-specific tricks.

The third block sits outside an editor's control entirely: certain page types are systematically excluded from Discover, regardless of quality. Job postings, petitions and forms, code repositories, and satire that isn't labeled as such are not eligible for the feed. Publishing that kind of content while expecting Discover traffic is simply wasted effort.

A common mistake also shows up in how the featured image gets implemented technically. Many CMS platforms set an og:image tag automatically, but not automatically at the right resolution: a theme that serves a 600-pixel-wide thumbnail for social previews technically has the tag in place, but misses the 1200-pixel minimum width Google specifies for Discover. Checking only whether an og:image tag exists, without checking the actual pixel width it delivers, misses the part that actually matters for Discover.

Google explicitly describes Discover traffic as less predictable and less dependable than traffic from keyword-driven search, and gives concrete reasons for that. A user's interests shift from one day to the next, which changes the pool of people an article is relevant to without warning. Google periodically adjusts which content types it shows in the feed, independent of any change made to a specific page. And general Search updates, the same ones that affect regular rankings, also ripple into Discover, sometimes more visibly there than in classic search results.

That explains a pattern that regularly confuses site owners: an article that had been generating thousands of visits from Discover for weeks suddenly disappears from the feed, with nobody having touched the page, no warning in Search Console, and no movement in the rest of the site's rankings. The instinctive read is usually "Google penalized us," and that's almost never what happened. More often, the article simply stopped matching the interest signals of the moment, or Google temporarily started showing less of that content type.

That's why Google frames Discover traffic as supplemental to keyword-driven search traffic, never as a channel to build a stable traffic forecast on. A business that depends on Discover for a meaningful share of its visits is leaning on the least predictable organic traffic source there is, with no dashboard that flags a drop before it happens.

In practical terms, that changes how the data should be read. The Discover report in Search Console deserves separate treatment from the regular search performance report, not the same cause-and-effect assumptions. A drop in search clicks can usually be tied to something concrete, like a ranking loss for specific keywords or a technical issue. A drop in Discover clicks often has no identifiable cause on the site's side at all, because the demand side, user interest, changed, not the supply side.

• Publish the featured image at least 1200 pixels wide, above 300,000 total pixels of resolution, and at a 16:9 aspect ratio, specified via the og:image tag or schema.org markup.
• Write headlines that describe the content accurately, without clickbait tactics or promises the article doesn't deliver on.
• Prioritize topics with genuine timeliness or a perspective that stays useful weeks later, instead of generic content with no point of view.
• Strengthen the signals behind helpful, people-first content that also underpin E-E-A-T: identifiable authorship, verifiable information, transparency about who's publishing.
• Take care of overall page quality, including Core Web Vitals, since it's part of the quality criteria Google associates with Discover-worthy content.
• Confirm the page is properly indexed and free of crawling restrictions, the minimum requirement before any other factor matters.
• Check the actual pixel width your CMS delivers for og:image, not just whether the tag is technically present.

• Treating Discover as a channel you can target with a specific keyword, when the system doesn't run on search queries at all.
• Assuming that meeting the image and content requirements guarantees appearing in the feed: Google is explicit that eligibility doesn't mean appearance.
• Reading a sudden Discover traffic drop as a penalty or a technical failure without first checking whether the rest of the site's rankings actually moved.
• Using a logo or a decorative image as the featured image instead of a high-resolution one tied to the article's actual content.
• Ignoring the 16:9 aspect ratio and assuming any large image is enough for eligibility.
• Reading the Discover report in Search Console with the same cause-and-effect expectations as the search report, when a Discover drop often has no identifiable cause on the site's side.
• Building a meaningful share of a business model on Discover traffic and treating it as a stable channel instead of a supplement to search traffic.