How Google Ranks Pages According to Patents Vintage
This guide synthesizes Bill Slawski's patent analyses into a structured overview of the ranking signals Google has patented. This is not speculation — these are signals described in granted patents and patent applications filed by Google engineers.
The Ranking Signal Taxonomy
SOP: How Google Ranks a Page
Step 1: Query Processing
When a user submits a query, Google first processes the query itself:
- Spelling correction — Fix obvious typos
- Query classification — Determine intent (informational, navigational, transactional, local)
- Entity recognition — Identify known entities in the query
- Query expansion — Add synonyms, related terms
- Data source selection — Route to appropriate index (web, news, images, local, shopping)
Step 2: Initial Retrieval
Google retrieves candidate documents from its index:
- Inverted index lookup — Find all documents containing query terms
- Phrase matching — Identify documents with query phrases (not just individual words)
- Semantic matching — Find documents matching query meaning via context vectors
- Entity matching — Find documents about the entity referenced in the query
Step 3: Scoring
Each candidate document is scored on multiple dimensions:
Step 4: Reranking
After initial scoring, results may be reranked based on:
- Diversity — Ensuring variety in results for ambiguous queries
- Personalization — Adjusting based on user history
- Freshness boost — Promoting recent content for time-sensitive queries
- De-duplication — Removing substantially similar results
- Blending — Inserting images, videos, news, maps into web results
Major Ranking Signal Categories
1. Content Signals
| Signal | Description | Patent Source |
|---|---|---|
| Keywords in titles | Title tag relevance to query | Ranking Signals (2018) |
| Keywords in headings | H1-H6 heading relevance | Ranking Signals (2018) |
| Related phrases | Co-occurring terms that establish topical coverage | Phrase-Based Indexing |
| Content depth | Comprehensive coverage of a topic | Quality Score Patent |
| Context terms | Words that establish the context/meaning of content | Context Vectors (2016) |
| N-gram language models | Natural language patterns in content | Ranking Signals (2018) |
2. Link Signals
| Signal | Description | Patent Source |
|---|---|---|
| Link authority (PageRank) | Quality and quantity of inbound links | PageRank Patent |
| Anchor text relevance | Descriptive link text matching query | Anchor Text Patents |
| Link position | Where the link appears on the page | Reasonable Surfer |
| Link context | Surrounding text of the link | Reasonable Surfer |
| Link velocity | Rate of link acquisition over time | Historical Data Patent |
| Co-citation | Being linked alongside authoritative sources | Co-Citation (2012) |
3. User Behavior Signals
| Signal | Description | Patent Source |
|---|---|---|
| Click-through rate | Percentage of searchers clicking a result | User Behavior (2011) |
| Dwell time | Time spent on page after clicking | User Behavior (2011) |
| Pogo-sticking | Returning to search results quickly | User Behavior (2011) |
| Query refinement | Searching again after visiting a result | User Behavior (2011) |
| Page watch time | Time spent consuming page content | Ranking Signals (2018) |
| Site blocking | Users choosing to block a site | Panda Integration |
4. Quality Signals
| Signal | Description | Patent Source |
|---|---|---|
| Quality score | Algorithmic quality assessment | Panda Patent (2011) |
| Content uniqueness | Non-duplicate, original content | Quality Score Patent |
| Ad-to-content ratio | Amount of advertising vs. useful content | Quality Score Patent |
| Template detection | Identifying auto-generated content | Quality Score Patent |
| Expert evaluation alignment | Matching quality rater assessments | Panda Patent |
5. Freshness Signals
| Signal | Description | Patent Source |
|---|---|---|
| Document age | When the content was first published | Historical Data Patent |
| Content update frequency | How often content is modified | Historical Data Patent |
| Link freshness | When links to the page were created | Historical Data Patent |
| Query deserves freshness | Whether the query expects recent results | Query Classification |
6. Entity Signals
| Signal | Description | Patent Source |
|---|---|---|
| Entity recognition | Content about known entities | Knowledge Graph Patents |
| Entity completeness | Coverage of entity attributes | Entity Patents |
| Entity relationships | Connections to related entities | Co-occurrence Analysis |
| Structured data | Schema markup providing entity data | Structured Data Patents |
Deep Dive Pages
- Panda & Quality Scores — How Google measures page and site quality
- Freshness & Historical Data — How document age, changes, and link history affect ranking
Key Takeaways
- No single signal dominates — Ranking is always a combination of multiple weighted signals
- Signal weights change per query — A news query weights freshness heavily; a research query weights authority
- User behavior is a real feedback loop — Click patterns and dwell time do influence rankings
- Quality is algorithmically measured — Google has specific, trainable systems for scoring content quality
- Entities complement keywords — Content that maps to known entities gets additional ranking opportunities
- History creates context — Your site's historical patterns form a baseline against which current signals are evaluated