Discover Feed #
The vertical scroll of cards that appears on the Google app, the Chrome new-tab page on Android, and Pixel home screens. It is generated per-user from interest signals — there is no single global ranking.
Surface & UX
5 termsCard / Card unit #
The atomic building block of the feed. A standard card shows a hero image (16:9 or 4:3), the title, the source name, and a freshness signal. Larger formats exist ("big card") for breaking news.
News tab #
Separate surface inside Google Search that aggregates news content. Eligibility overlaps with Discover but is not identical — News favors recency and original reporting.
Sticky session #
When Discover detects strong engagement on a topic, it temporarily over-weights related cards in the same browsing session. This is the mechanism behind the typical "Discover spike".
Topic chip #
The small interest pills users tap to refine their feed (e.g. "AI", "Premier League"). Subscribing to a chip pushes related publishers up the user's personal ranking.
Discoverability
8 termsEligibility #
The minimum technical bar a URL must clear to be considered for Discover: indexable in Google Search, mobile-friendly, served over HTTPS, and free of policy violations. Eligibility ≠ ranking — it just opens the door.
Web Profile #
Public Google-hosted page at profile.google.com/cp/<ID> that aggregates a publisher's signals (logo, social links, last 30 days on Discover). The presence of this URL is the cleanest existence proof that a domain is recognized as a Discover entity.
Indexing API #
Google's push-indexing endpoint. Officially restricted to job postings and live-stream pages, but news publishers often use it as a fast-lane for breaking content. Use carefully — abuse triggers throttling.
Publisher Center #
Google's dashboard at publishercenter.google.com where publishers manage their logo, sections and Following-button identity. Mandatory for News-tab inclusion, optional but recommended for Discover.
Sitemap #
An XML list of canonical URLs the publisher wants indexed. For news, a dedicated news.xml with <news:publication> blocks accelerates crawl of articles less than 48 h old.
Robots meta directives #
<meta name="robots"> tags that instruct Googlebot. The combo index, follow, max-image-preview:large, max-snippet:-1 is the de-facto Discover-eligibility recipe.
max-image-preview:large #
The robots-meta value that grants Google permission to display large image previews. Without it, your hero image will not be used as a Discover card thumbnail. Single most under-set tag in eligibility audits.
IndexNow #
Open protocol (Bing, Yandex, Cloudflare) that pings search engines when a URL is published or updated. Google does not officially support it, but many CDNs flip it on by default — neutral for Discover but useful for general SEO.
Structured data
5 termsJSON-LD #
The recommended format for declaring schema.org markup on a page. A single <script type="application/ld+json"> block in <head> can carry multiple typed entities (Article, BreadcrumbList, FAQPage, …).
Article schema #
Schema.org type for editorial content. Required fields for Discover-grade markup: headline, image, datePublished, author, publisher.
NewsArticle schema #
Stricter sub-type of Article reserved for time-sensitive news. Adds dateline and recognizes printSection. Misusing it on evergreen content can trigger demotions.
BreadcrumbList #
Schema declaration of the page's hierarchical position. Powers the breadcrumb that sometimes replaces the URL line in cards and search snippets.
ImageObject #
Schema sub-entity to declare an image with explicit width, height and url. Strongly recommended inside Article markup so Google can pick the right hero crop.
Content quality
6 termsE-E-A-T #
Experience, Expertise, Authoritativeness, Trustworthiness — the four-letter framework Google's quality raters use to evaluate content. Heavily weighted on YMYL topics.
YMYL — Your Money or Your Life #
Topics that materially affect the reader's wellbeing: health, finance, legal, safety, parenting, civics. Discover applies an extra layer of scrutiny here, including stricter author-page expectations.
Helpful Content System #
Google's system (rolled into the core algorithm in 2024) that demotes sites whose content is written primarily for search engines rather than humans. A site-wide signal — one bad section can drag the rest down.
Site Reputation Abuse #
Policy targeting third-party content (often affiliate or coupon) hosted on a high-authority domain to siphon its rankings. Active since May 2024 and aggressively enforced via manual actions.
Spam policies #
The catalogue of behaviors that get content removed from Discover entirely: cloaking, sneaky redirects, doorway pages, scraped content, hidden text, hacked content, automatically generated content lacking value.
Cloaking #
Showing different content to Googlebot than to human visitors. Hard violation — triggers manual action and full Discover suspension.
Image
5 termsHero / Lead image #
The top visual of an article, also called the lead image. This is the asset Google selects (in 92% of cases) for the Discover card thumbnail.
Aspect ratio #
The width-to-height proportion. Discover crops cards to either 16:9 (mostly) or 4:3. Anything below 1200 px wide gets pillar-boxed and ranks worse.
Image dimensions (1200×800+) #
Google's official guidance: hero images must be at least 1200 px on the longest side. Below that, the image becomes ineligible to render at full size in Discover, even if everything else is correct.
AI image disclosure #
C2PA / IPTC metadata that flags an image as AI-generated. Discover does not penalize AI imagery per se but expects editorial visuals to be disclosed when they are not photographic evidence.
Image alt text #
The alt="" attribute. Used for accessibility and as a tie-breaker when Google has to choose between competing candidates for the card thumbnail. Keep it descriptive, not keyword-stuffed.
Title & metadata
5 termsH1 #
The single first-level heading on the page. It should match the title users see in the Discover card — divergent H1 vs Title leaks ranking signals and confuses Google's title-rewriting layer.
Title tag #
The <title> in <head>. Discover often shows it verbatim, sometimes truncated to 70 characters. The single highest-leverage piece of copy on the entire page.
Meta description #
The 150–160 character summary in <head>. Discover only renders it on a fraction of cards but it remains a tie-breaker for click-through prediction.
Open Graph #
The og:* meta family that controls how the URL renders when shared on social. Discover falls back to og:image when no <link rel=image_src> or schema image is present.
Canonical URL #
The single authoritative URL for a piece of content, declared via <link rel="canonical">. Discover uses it to consolidate signals across duplicate or syndicated copies.
Engagement signals
5 termsCTR — Click-through rate #
Clicks ÷ impressions. Discover's primary feedback signal. A 5% CTR is the rough median; sustained <2% suppresses future distribution; 10%+ unlocks sticky sessions.
Dwell time #
Time elapsed between a click and the user returning to the feed. Long dwell = signal that the article matched the promise of the title. Short dwell = the title oversold.
Pogo-sticking #
When a user clicks a card, returns to the feed within seconds, then clicks a different card. The dual signal (short dwell + immediate alternative) is the most damaging engagement pattern.
Long click vs short click #
Internal Google distinction. A long click ends the user's task (no further query); a short click sends them back to search. Discover uses the same distinction adapted to the feed.
Bounce-back #
User-event Discover instruments specifically: tap card → arrive on article → press back → return to feed. The latency between "arrive" and "back" is one of the cleanest quality proxies available.
Targeting & personalization
4 termsKnowledge Graph #
Google's structured database of entities (people, places, organizations, concepts). A topic that maps cleanly to a Knowledge Graph entity surfaces faster in Discover than free-form keywords.
Topical authority #
Site-level credibility on a defined topic cluster. Built by depth (multiple in-depth articles), recency (sustained publication cadence), and citations (links from other authorities).
Web & App Activity #
The user-account setting (myactivity.google.com) that gates personalization. Disabled = the user sees a generic, geo-only Discover; enabled = full personal feed.
Personalization #
The user-specific re-ranking layer applied on top of the global ranking. Inputs: search history, site-following, topic-chip subscriptions, time-of-day patterns, language, location.
Performance metrics
5 termsImpression #
Counted when a card enters the viewport for at least one second — strict definition vs. Search where impression = position in result page. A scroll-past does not count.
Click #
Tap on the card thumbnail or title that opens the article. Multiple clicks within a session de-duplicate to one in Search Console.
Traffic spike #
The non-linear traffic burst typical of Discover — easily 10–100× a publisher's organic baseline. Rare per article (~5% of eligible URLs trigger one), volatile, and impossible to forecast.
Decay / Half-life #
An article's traffic typically halves every ~6 hours after the first impression burst. Half-life varies by topic — evergreen guides decay slowest, news decays fastest.
Freshness / Recency #
Boost given to articles published or substantively updated in the last 24–48 h. Updating dateModified without changing real content is detected and demoted.
Penalties & demotions
4 termsDemotion #
An algorithmic downgrade applied at the URL, section, or domain level. Different from removal — content remains indexable but loses Discover distribution.
Manual action #
Human-reviewed penalty visible in Search Console. Triggered by clear policy violations (spam, hidden text, hacked content, site-reputation abuse). Requires a reconsideration request to lift.
Algorithmic suppression #
Silent demotion not reported in Search Console. Detectable only by sudden, sustained drops in Discover impressions while Search remains stable.
Core / Spam update #
Periodic algorithm refreshes (3-6 per year). Core updates re-tune ranking; Spam updates target policy-violators specifically. Discover impact is often more violent than Search impact.
Tools & dashboards
3 termsSearch Console #
Free Google dashboard at search.google.com/search-console. The only authoritative source of impressions, clicks and CTR for both Search and Discover.
Discover performance report #
Search Console's dedicated Discover view. Only appears once your domain has accumulated >100 Discover impressions in 16 months — its mere existence is a signal that you are eligible.
Publisher Center #
Google's dashboard at publishercenter.google.com. Lets publishers manage their logo, sections and Following-button identity. Optional but recommended for Discover.