Google Discover can feel random. One day traffic spikes. The next, it disappears. Many publishers assume ranking is the problem. New research shows most failures happen much earlier.
SDK-level analysis now reveals how Discover really works. Content passes through strict eligibility checks before ranking even begins. Blocking, image rules, freshness decay, and predicted engagement decide visibility first.
This breakdown explains the full Discover pipeline. You’ll see why pages never appear, how ranking actually works, and what to fix before optimization even matters.
What This Research Analyzed — And Why It Matters
SDK-Level Visibility Into Google Discover
The research was conducted by :contentReference[oaicite:0]{index=0}. It analyzed observable signals inside the Discover app framework.
This is not guesswork or correlation-based advice. The findings reflect system-level behavior visible before ranking decisions occur.
Why This Is Different From Typical SEO Advice
Most Discover tips focus on headlines and images. This research exposes what happens before those elements even compete.
It explains why strong content can fail without ever entering ranking.
The Google Discover Pipeline
Stage 1–3: Crawl, Parse, Classify
Discover first crawls and understands the page. It reads core meta tags like title and image.
The system then classifies content type. News, evergreen, and mixed formats follow different paths.
Stage 4: Publisher-Level Blocking
This is the most critical gate. Blocking happens before ranking.
One “Don’t show content from this site” action suppresses the entire domain for that user. There is no positive override.
If you are blocked, you cannot rank.
Stage 5–9: Matching, Ranking, Delivery, Feedback
Only unblocked content reaches interest matching. Google predicts engagement using server-side models.
The feed layout is built, content is delivered, and user behavior is logged instantly.
The Most Important Finding: Blocking Comes Before Ranking
Why Publisher Blocking Is So Powerful
Blocking is sitewide and permanent at the user level. Publishers receive no notification.
There is no recovery mechanism beyond rebuilding trust with new users.
What Triggers Blocks
Common triggers include clickbait fatigue, repetitive topics, and low perceived value.
Mismatch between headline promise and content delivery accelerates blocks.
How Google Discover Actually Ranks Content
The pCTR Model Explained
Discover uses predicted click-through rate before impressions occur.
This server-side pCTR model estimates how likely a user is to click if shown the card.
Signals Fed Into the Model
- Page title from og:title
- Image size and quality
- Content freshness
- Historical clicks and impressions
- Image load success
Freshness Decay Windows
Discover’s Time-Based Boost System
Freshness follows strict timelines.
- 1–7 days: strongest visibility
- 8–14 days: moderate exposure
- 15–30 days: limited reach
- 30+ days: steady decay
Evergreen Is the Exception
Evergreen content follows a different path. It must earn sustained engagement to remain visible.
Most pages never qualify.
Image Requirements That Decide Visibility
No Image Means No Discover Card
An og:image is mandatory. Without it, the page cannot appear.
Discover evaluates six page-level meta tags before rendering a card.
1200px Images Control Card Size
Images must be at least 1200px wide for large cards.
Smaller images are shown as thumbnails and earn fewer clicks.
Meta Tag Fallback Logic
If og:title is missing, Discover falls back to Twitter title, then HTML title.
Missing tags reduce reliability and pCTR confidence.
Two Meta Tags That Can Kill Discover Eligibility
The following tags can block Discover entry entirely:
- nopagereadaloud
- notranslate
Either tag can prevent feed inclusion before ranking begins.
Personalization Layers Inside Discover
Interest Graph Signals
Discover uses search history, app usage, and topic engagement.
Publisher Signals
Publisher Center registration, consistency, and historical trust influence exposure.
Individual User Actions
Follows, saves, dismissals, and time on article matter.
URL-level dismissals are permanent for that user.
Google Discover Is Constantly Experimenting
Scale of Experimentation
Roughly 150 server-side experiments can run at once.
Over 50 display features affect card appearance.
Why Feeds Differ So Much
Users are placed into different experiment buckets.
The same content can perform very differently.
Real-Time Feed Mutation
Discover updates without refresh.
Cards can be added, removed, or reordered while scrolling.
Why Pages Never Reach Ranking
- Publisher-level block
- Missing or broken images
- Meta tag failures
- Freshness window missed
- Weak historical engagement
- User-level dismissals
How to Optimize for Google Discover
Eligibility First, Ranking Second
Fix technical blockers before chasing clicks.
Audit meta tags and image compliance.
Reduce Publisher-Level Risk
Avoid clickbait. Improve topic diversity.
Focus on value per impression.
Optimize for pCTR Signals
Use clear titles, strong images, and consistent publishing.
What This Means for SEO Strategy in 2026
Discover is not bonus traffic. It is a separate system.
Eligibility matters more than optimization tricks. Trust compounds. Penalties linger.
Key Takeaways
- Discover can block you before ranking
- Blocking is permanent at user level
- Images and freshness dominate visibility
- pCTR predicts success before impressions
- Experiments make outcomes unpredictable
FAQs
Why doesn’t my site appear in Google Discover?
Your site may be blocked, missing required images, or failing freshness and engagement thresholds.
How does Google Discover rank content?
It uses predicted click-through rate, freshness, images, and historical engagement.
What image size is required for Google Discover?
Images should be at least 1200px wide to qualify for large cards.
What is pCTR in Google Discover?
pCTR is a server-side prediction of how likely a user is to click a card.
Can Google Discover block an entire site?
Yes. One user action can suppress a full domain for that user.
How long does content stay eligible for Discover?
Most content peaks within seven days, then declines rapidly.
Related reading: previous analysis on Google’s expanding content surfaces.
Need help fixing Discover visibility? Explore our Affordable SEO Services.
