Google Discover Optimization: Earning Traffic From a Feed You Can't Query

Most of your search strategy assumes a query exists: someone types something, Google ranks pages, you compete for position. Discover breaks that model entirely. There is no query, no SERP, and no keyword to target. Instead, Google decides what to show each user based on their inferred interests, and your content either matches that interest graph or it doesn't. Optimizing for it means thinking like a recommendation engine, not a search index.

How the Feed Actually Decides What to Surface

Discover is a personalized content feed on the Google app, the mobile homepage, and Chrome's new-tab on Android. It pulls from Google's regular web index but ranks by predicted interest rather than relevance to a search string. The system maps each user to a set of entities and topics they engage with, then matches that profile against fresh, indexed content associated with the same entities.

The practical consequences:

• You cannot trigger it on demand. There is no input you control that guarantees a surface. You can only make content eligible and attractive.
• It is interest-driven, not intent-driven. A user isn't looking for an answer; they're browsing. Content that satisfies a known need (a how-to for an active problem) often underperforms content that sparks latent curiosity.
• Freshness and timeliness weight heavily. Most Discover traffic flows to content published or meaningfully updated recently, though evergreen pieces tied to durable interests do recur.
• It is overwhelmingly mobile. Design and image decisions should assume a phone screen.

Eligibility: The Non-Negotiable Baseline

Before any optimization matters, content has to be eligible. Google has stated that Discover follows its core content policies and that there is no markup or feed submission to opt in. Confirm these are in place:

• The page is indexed. If it's not in Google's index, it can't appear in Discover. Verify in Search Console's URL Inspection tool.
• You are not blocking the image. Large images are the single biggest Discover lever, but they require explicit permission. Add max-image-preview:large to your robots meta tag:
  
  <meta name="robots" content="max-image-preview:large">
• Content meets the news and quality policies. Medical misinformation, manipulative or clickbait behavior, and adult content are disqualifiers even if otherwise high quality.

Note that Discover performance has its own report in Search Console, but it only appears once your site has accumulated meaningful Discover impressions. No report means no traffic yet, not a configuration error.

Engineer the Image First

In a feed, the image is the headline. A small thumbnail and a large hero image are categorically different experiences, and Google's own guidance is explicit that large, high-quality images drive significantly more engagement than thumbnails.

• Serve images at least 1200px wide. Below that threshold Google may render a small thumbnail or omit the large preview entirely. This is the most common reason a page never gains traction.
• Choose images that read at a glance on a phone. Avoid logos, text overlays, and busy compositions. A single clear subject with strong contrast wins.
• Make the image earn the click honestly. A misleading or sensational image is a manipulative-behavior policy risk and trains the system that your content disappoints.
• Use a compelling primary image per article, not a generic site banner. The feed shows one image; treat it as the cover of that specific piece.

Titles That Earn the Tap Without Baiting It

Discover titles operate under a peculiar tension: they must be intriguing enough to interrupt a scroll, yet Google actively suppresses clickbait, exaggeration, and curiosity gaps that the content doesn't pay off. The winning register is specific and confident, not breathless.

• Lead with the concrete payoff or the surprising fact, not a withheld secret. "The setting most people miss" baits; "Three settings that cut battery drain" delivers.
• Avoid all-caps, excessive punctuation, and emotional manipulation. These are explicit clickbait signals.
• Don't keyword-stuff the title. There's no query to match, so optimizing the title for a search phrase is wasted effort and reads as robotic.
• Keep the page title and the on-page H1 aligned with the body. A mismatch between promise and delivery is the behavior Discover penalizes hardest.

Topical Authority Replaces Keyword Targeting

Because Discover matches users to entities and topics, the unit of optimization is not the page but the topic cluster your site is associated with. Google is more likely to surface content from a site it recognizes as authoritative on a subject the user follows.

1. Pick lanes and publish deeply within them. A site that covers one vertical thoroughly builds a clearer entity association than one that publishes scattered content across many topics.
2. Cover entities, not just keywords. Name people, products, places, and organizations precisely and consistently. The feed reasons about entities; help it map your content to them.
3. Demonstrate first-hand experience and expertise (E-E-A-T). Clear authorship, author pages with credentials, and original reporting or analysis all reinforce the trust signals Discover leans on.
4. Interlink within the cluster. Internal links that connect related articles reinforce the topical relationship and keep readers in your ecosystem after a Discover entry.

Timeliness and the Trend Window

Discover rewards content that aligns with rising interest. The strongest single tactic for predictable traffic is publishing useful, well-imaged content into a topic that is actively trending within your authority lane. This doesn't mean chasing every news cycle; it means having enough topical depth that when interest in your subject spikes, you have credible, fresh content ready to be matched. Updating a strong evergreen piece with a current angle and a fresh publish-meaningful timestamp can also re-trigger eligibility.

Common Mistakes

• Treating Discover like SEO. Optimizing titles for keywords, chasing backlinks for a target phrase, and obsessing over query intent all miss the point. There is no query.
• Shipping 800px images. The most common silent failure. Without a 1200px+ image and max-image-preview:large, you forfeit the feed's primary engagement driver.
• Reading volatility as a penalty. Discover traffic is spiky by nature; a piece can drive a surge for days then vanish. Sharp drops are usually the feed moving on, not a manual action.
• Clickbait that converts once. Misleading titles and images may win an initial tap but degrade the system's trust in your domain, suppressing future surfaces.
• Expecting instant results. The interest-matching model needs your site to build entity associations and engagement history before it surfaces you reliably.

How to Measure What You Can't Query

Use the dedicated Discover report in Search Console (separate from the Search results report) to track impressions, clicks, and CTR over time. Since there are no keywords, analyze by page and date: identify which articles and which topics earned surfaces, then double down on those clusters. CTR in Discover is largely a referendum on your image-and-title pairing, so treat a low CTR on high impressions as an image or headline problem, not a content problem. Over time, the pages that recur in the report tell you exactly which entities Google has learned to associate with your domain, which is the closest thing to a keyword map this channel offers.

Discover rewards a fundamentally different discipline than search: build genuine authority in defined topics, package each piece with a large, honest image and a specific title, keep your content fresh and policy-clean, and let the recommendation engine match you to the people already interested. You're not ranking for a query. You're earning a place in someone's feed.