A publisher opens analytics on a Tuesday and finds 38,000 sessions pouring in from a source labeled only Google / Discover. One article, published three weeks earlier and mostly forgotten, is carrying all of it. By Thursday the surge is gone. Nobody changed a setting to make either thing happen.
That sequence confuses people because it breaks the model search trains into us. A query goes in, a ranked page comes back, and the same query returns roughly the same result tomorrow. Discover runs on the opposite logic. No one typed anything, and the feed moved on for reasons that never appear in a keyword report.
This guide covers what Google Discover is, what makes a page eligible, what the feed appears to reward, and where a publisher’s control actually ends. That last part is shorter than most people hope, which is exactly why it matters where the effort goes.
Discover is a feed, not a search result #
Discover is the personalized stream inside the Google app and on the Chrome home screen on Android. It surfaces articles, videos, and pages based on what Google infers a person cares about, drawn from signals like past searches, YouTube activity, and which cards someone taps or scrolls past.
The defining trait is simple. It is queryless.
No search box sits in front of the feed, and that single fact rewrites the whole problem. Keyword targeting, search intent matching, and SERP position, the levers that decide whether a page wins a query, have nothing to act on where no query exists. A page earns its place by being what a specific person wants to read next, something Google predicts rather than retrieves in response to a request.
What actually makes a page eligible #
Google’s own documentation is unusually blunt here: content is automatically eligible to appear in Discover if it is indexed and meets Discover’s content policies, and no special tags or structured data are required to get in.
That kills a popular assumption. There is no Discover schema, no feed to submit, no opt-in.
Eligibility, though, is a low and slightly misleading bar, because Google also states plainly that being eligible to appear in Discover is not a guarantee of appearing. Every indexed page that follows the policies sits in the pool, and the feed then selects from that pool using personalization and quality signals no publisher can see. The work was never getting eligible. It is becoming a page Google is willing to recommend to someone who never asked for it.
The image rule that quietly gates the card #
Discover cards are visual objects, and the image, more than the copy, is what gets a card tapped instead of scrolled past. Google is specific about what counts as large enough.
Three conditions line up. The image needs to be at least 1200 pixels wide, carry more than 300,000 total pixels, and sit near a 16:9 aspect ratio, with the large preview switched on either by the max-image-preview:large robots directive or by serving AMP.
The quiet failure is almost never the photograph. A site can ship a sharp 1600-pixel image and still render a cramped thumbnail in the feed, because a plugin or a conservative default set max-image-preview to standard across the entire domain. A single line in the page head decides whether the feed is even allowed to use the large version:
<meta name="robots" content="max-image-preview:large">That line is worth confirming in the rendered source before touching anything else, because if it is missing or set to standard, the most persuasive part of the card is switched off long before any content question comes up.
What the feed rewards, and what it drops #
Because Discover pushes content at people who did not request it, Google weighs heavily on whether a source is trustworthy enough to recommend unprompted. Its documentation and its February 2026 core update, the first ever aimed only at Discover, point the same way: original, in-depth, timely work from credible publishers is favored, while clickbait, exaggerated previews, and sensationalism that trades on morbid curiosity get filtered down.
The split is consistent enough to plan around.
| Discover appears to favor | Discover appears to suppress |
|---|---|
| Original reporting or analysis a reader cannot get from ten near-identical pages | Summaries that repackage what everyone else already published |
| Titles that capture the actual essence of the page | Curiosity-gap headlines that promise more than the body delivers |
| Visible expertise and named authorship, the E-E-A-T signals | Anonymous content on topics that clearly need credibility |
| A timely reason for the page to surface right now | Thin filler with no particular reason to appear today |
None of these are switches. They are the qualities the system was tuned, in early 2026, to prefer, and a page missing several of them can sit in the eligible pool without ever being shown.
Size is not the gate here, though it is easy to assume it is. A small independent site publishing genuinely original work on a subject it can speak to with authority sits in the same pool as a national outlet, and the February update’s emphasis on credible publishers rewards demonstrated credibility rather than raw traffic. The disadvantage a small site carries is not exclusion, it is that trust signals take longer to establish.
Titles: where a strong headline turns into clickbait #
Discover is unforgiving about titles in a way ordinary search is not, because the headline and the image are all a person sees before deciding. Google’s guidance is to use titles that capture the essence of the content and to avoid withholding the crucial information a reader needs to understand what the page is about.
The line is thinner than it sounds. A vivid headline is fine.
A headline that hides the payoff to force the tap is the exact pattern the content policies name as misleading preview content, and the February 2026 update sharpened the response to it. Consider two versions of the same story. “The pricing change that cost a SaaS company 12% of signups” tells the reader what they are getting and still earns the click on merit. “This one pricing mistake is quietly killing your signups” withholds the substance and inflates the stakes, and the second kind is what Discover’s own policy language targets.
What you can control, and what you cannot #
Most Discover frustration comes from spending effort on the second half of this list. The honest division looks like this:
- Controllable: indexation, the
max-image-preview:largedirective, image dimensions, title and preview honesty, demonstrated expertise and authorship, and staying inside the content policies. - Not controllable: who the feed decides is interested, when a page surfaces, how long a surge lasts, and the personalization signals that shift from one reader to the next.
That asymmetry is the strategy. A publisher can make a page fully qualified and genuinely worth recommending, and then the timing and the audience belong to Google. Treating a Discover spike as something to reproduce on demand leads to chasing patterns that were never in anyone’s hands, while treating it as a reward for durable quality keeps the effort on the parts that answer to effort.
Read the Discover report before drawing conclusions #
Search Console adds a Discover performance report, but only after a site has actually received Discover traffic, so a brand-new or low-traffic site will not see it yet. When it appears, it shows impressions, clicks, and click-through rate for the feed specifically, separated from ordinary search.
That separation is the point. It is the one place the queryless surge becomes legible.
Reading it changes the questions worth asking. A page that pulls a large surge and then goes quiet has not necessarily been penalized. The more likely event is that the personalized window it fit has closed, which the report shows as impressions dropping while the page itself is unchanged, and telling that apart from a genuine quality problem is nearly impossible from analytics alone.
Discover rewards being the kind of source Google will vouch for without being asked, and almost nothing about it answers to the tactics that move a search ranking. The durable move is to make each page genuinely worth recommending, confirm the image directive and the honesty of the title, and then read the Discover report to learn which of your work the feed already trusts, rather than trying to force a surface built to decide on its own.