Discover feed

Long definition

Discover is a no-query surface. Users scroll a feed of articles, videos, and short-form content selected by Google based on search history, app activity, location, and explicit topic follows. Each card shows a large image, headline, publisher, and timestamp. Tapping the card sends the user to the source URL — these are real visits, not previews like image-pack clicks.

Eligibility is harder to engineer than for SERP features. Google has been explicit: there's no Discover-specific markup, no submission flow, no Publisher Center hook. The signals that correlate are clean E-E-A-T, large compelling images (1200px wide minimum, with max-image-preview:large in robots meta), a strong topical track record, and content the algorithm reads as freshly relevant to a known interest. Evergreen content can perform — Discover doesn't require news velocity — but timeliness helps.

Search Console exposes a Discover performance report only after your site has accumulated Discover impressions. New sites won't see the report until Google starts surfacing them. The report shows impressions, clicks, and CTR per URL, but no query data — there is no query.

Volatility is the defining feature. A site that gets 50,000 Discover sessions in October can get 500 in November with no obvious change. Discover traffic should be treated as bonus, not baseline. Building a business model on Discover-dependent traffic has burned many publishers.

YMYL pressure applies. Health, finance, and political content goes through harder filters in Discover than in regular Search. A site that ranks fine organically but gets demonetized claims wrong can find Discover impressions collapse with no Search ranking change.

Common misconceptions

• "AMP is required for Discover." Removed the same time AMP requirements dropped from Top Stories. Fast-loading non-AMP pages compete equally.
• "Adding max-image-preview:large opts me into Discover." It makes large preview images possible — without it, your card shows a tiny thumbnail and CTR collapses. But the directive doesn't trigger inclusion. Inclusion is algorithmic.
• "Discover impressions count toward my Search ranking." They're a separate surface. Discover engagement signals likely feed broader quality models, but the Search Console Discover report is isolated from the Search report.
• "You can submit articles to Discover." No submission path exists. No Publisher Center flow, no API, no schema. Discover surfaces what its algorithm decides is worth surfacing.