Personalization layer over the structured entity page. Content selection and ordering shape to the viewing user, so two users see different facts, sections, and emphasis on the same entity in line with their individual interests and context.
Patent Overview
- Inventor
- Jeromy William Henry, others
- Assignee
- Google LLC
- Filed
- 2019-12-19
- Granted
- 2021-02-09
- Application Number
- US 16/720,829
The Challenge
The Challenge
A static entity page shows the same content to every user. But what matters about an entity depends on who is looking: a fan wants tour dates, a researcher wants biography, a business contact wants leadership. Personalization tailors the page to viewer intent.
- Static Pages Serve Average User Poorly — A single static layout averaged across all users serves no specific user well. Each user has different reasons for visiting an entity page.
- User Context Reveals Likely Interest — Recent queries, location, time, prior entity engagement all hint at why the user is visiting. The signal is rich if the system reads it.
- Personalization Must Preserve Identity — Core entity facts (name, type, primary description) must appear for everyone. Personalization affects secondary content, not the entity's fundamental identity.
- Privacy Constraints Bound Personalization — Some signals are off-limits for personalization. The system respects privacy controls and sensitive-category boundaries.
- User Must Recognize The Same Entity — Two users on the same entity page should both recognize it as the same entity. Personalization cannot make the page feel like two different entities.
Innovation
How The System Works
The system reads per-user context signals, scores each candidate content item on relevance to the user, applies per-user weights to section ordering and content selection within sections, renders a personalized variant of the entity page, and refines personalization from interaction feedback.
- Collect User Context — Session, location, time, profile, recent queries, prior entity engagement. Context signals feed personalization.
- Apply Privacy Boundaries — Per-user privacy settings filter which signals participate. Sensitive categories are excluded by default.
- Score Candidate Content — Per content item (fact, related entity, content source), score relevance to the user's context. Items aligned with user interests score high.
- Order Sections By User Relevance — Sections with highest-scoring content rise. A user interested in the entity's products sees products near the top; one interested in leadership sees leadership.
- Select Content Within Sections — Within each section, top-scoring items surface. The standard composition logic operates on the user-scored candidate set.
- Render Personalized Variant — Compose the page with user-specific ordering and selection. Core entity identity (name, type, primary description) remains constant; secondary content adapts.
- Capture Interaction Feedback — Per-user clicks, dwell, expansions on the personalized page feed back into personalization scoring. The system learns user preferences continuously.
Same Entity, Personalized View
The patent's load-bearing idea is to keep the entity constant while letting the lens through which the user views it personalize. Identity is shared; emphasis adapts.
Identity Stays, Emphasis Adapts
Core entity facts appear for everyone. The personalization layer shapes which secondary facts surface, which sections lead, and which related entities and content sources display.
- Per-User Context Scoring — Each user's context produces per-item relevance scores. The same entity page composes differently for different users.
- Section Reordering — Sections with user-relevant content rise. The page leads with what matters to this user.
- Privacy-Bounded — Privacy settings filter which signals participate. Sensitive categories are excluded.
Technical Foundation
Technical Foundation
The patent specifies the context aggregator, the privacy filter, the relevance scorer, the section orderer, the content selector, and the personalized renderer.
- Context Aggregator — Per user, aggregates session, location, time, profile, and history into a structured context vector for downstream scoring.
- Privacy Filter — Per-user privacy settings filter which signals reach personalization. Sensitive categories opt out by default.
- Relevance Scorer — Per candidate content item, scores relevance against the user context vector. Learned scoring model trained on engagement data.
- Section Orderer — Sections sort by aggregate scoring of their candidate items. Highest-relevance sections lead the page.
- Content Selector — Within each section, top-scoring candidates fill the slots. Standard composition logic with user-aware ordering.
- Personalized Renderer — Composes the page with per-user ordering and content. Core entity identity stays constant; personalized layers shape the rest.
The Process
The Process
Personalization runs in the entity-page composition path. Latency is bounded; context aggregation and scoring happen alongside the entity-record fetch.
- User Requests Entity Page — Authenticated request includes user identifier. Both entity-record fetch and context aggregation fire.
- Aggregate Context — Session, location, time, profile, and history aggregate into the context vector. Privacy filters apply.
- Score Content Items — Per candidate content item from the entity record, the relevance scorer computes a user-context score.
- Order Sections — Sections sort by their candidate-item scoring. User-relevant sections lead.
- Select Within Sections — Per section, top-scoring candidates surface. Composition logic respects template structure.
- Render Personalized Page — Personalized compositor produces the page. Core entity stays constant; sections and content adapt.
- Log Interactions — Per-user engagement logs feed personalization training. The system improves continuously.
Quality Control
Quality Control
Personalization that drifts produces wrong pages. The patent specifies safeguards.
- Core Identity Anchor — Name, type, primary description always appear. Personalization shapes the rest. Users always recognize the entity.
- Privacy Default Conservatism — Sensitive categories opt out by default. Personalization respects privacy boundaries.
- Bounded Reordering — Section ordering changes but core structure remains. Excessive reordering would confuse users; bounds prevent it.
- User Override — Users can disable personalization, view the standard layout, or reset their profile. First-class controls keep users in charge.
- Engagement Validation — Personalized variants are validated against engagement outcomes. If personalization makes engagement worse, the model retrains.
Real-World Application
Personalized entity pages appear in Google search products as logged-in users see entity surfaces tailored to their history. The patent's primitives underpin per-user entity experiences in Knowledge Graph-driven surfaces.
- Per-user Personalization Granularity — Each user sees a personalized variant of the entity page. Two users on the same entity see different ordering and content emphasis.
- Core-anchored Identity Stability — Core entity facts always appear. Users always recognize the entity regardless of personalization.
- Privacy-bounded Signal Scope — Privacy filters bound which signals contribute. Sensitive categories opt out by default.
Why Niche Audiences Compound Through Engagement
Personalization rewards alignment with user interests. Content aimed unmistakably at a specific audience earns repeat engagement from that audience, compounding visibility on personalized entity pages.
Why Logged-In Search Differs From Logged-Out
Logged-in users get personalized entity pages; logged-out users see the standard layout. Brands targeting logged-in audiences need to think about personalized signal alignment, not just SERP ranking.
<\/section>What This Means for SEO
What This Means for SEO
The patent personalizes the entity page so the same entity surfaces different facts, section order, and emphasis per viewer, refined by interaction feedback. SEO implication: for logged-in audiences, aligning your content with a specific audience's interests compounds visibility, since personalization rewards clear audience fit.
- Identity Stays, Emphasis Adapts — Core entity facts appear for everyone while secondary facts and section ordering shift per viewer. Ensure the facts that matter to your target audience are clearly associated with the entity, so personalization can elevate them for the right users.
- Clear Audience Targeting Wins Personalized Slots — Personalization rewards alignment with user interests. Content aimed unmistakably at a specific audience earns repeat engagement from that audience, compounding its placement on personalized entity pages over diffuse, everyone content.
- Logged-In Differs From Logged-Out — Logged-in users get personalized variants; logged-out users see the standard layout. Brands targeting logged-in audiences must think about personalized-signal alignment, not only their static SERP position.
- Interaction Feedback Tunes Emphasis — Personalization refines from interaction feedback. Content that earns clicks and engagement from a given audience strengthens its future emphasis for similar users. Early engagement from your target segment compounds.
- Serve The Right Facet Per Audience — Different viewers want different facets of one entity (tour dates, biography, leadership). Publishing facet-specific content that maps to each audience's intent lets the personalization layer pick your content for the matching viewer.
- Context Signals Drive Selection — The system reads per-user context to score candidate content. Content with clear contextual cues (audience, purpose, relationship to the entity) is easier to score and place for the matching user than ambiguous general content.
- Niche Specialization Beats Generality Here — Because emphasis adapts to the individual, deeply serving a niche audience earns stronger personalized placement with that audience than broad coverage earns with anyone. Depth for a defined segment is the lever.