The Entity Trust Patent
In May 2006, Ramanathan Guha — the man who invented RSS and helped architect Google's knowledge systems — filed a patent describing a search ranking system built on a premise the SEO industry has been misreading for two decades. The system doesn't measure trust through links. It measures trust between people. Experts label content, users express trust in those experts, and search results are re-ranked by the aggregated trust of the entities who vouched for the documents. This patent adds a second question on top of relevance: not only what is being said, but who is labeling, endorsing, or standing behind it. That's not to say links don't matter. They do. This patent describes a trust dimension that operates alongside the link graph, not instead of it.
Throughout this article, I use "people" as shorthand. Strictly speaking, the patent defines "entity" broadly: a person, group, organization, website, business, institution, government agency, or any other identifiable source. The distinction from classic link-graph trust is not pages linking to pages, but entities trusting and labeling entities and resources.
If you've read my analysis of the Seed Distance patent, that system measures how far your page is from the sites Google trusts through the link graph. This patent answers a different question entirely: how much do the people who vouch for your content get trusted by other people? Seed Distance is spatial. This patent is social. Together, they form two dimensions of the same trust architecture.
I had not read this specific patent before this analysis. I was aware of many of the principles behind it — I've written about entity trust in my TISEL essay, and I've been applying entity-anchored link building in my campaigns since 2016. But the full text of US7603350B1 is new to me. What struck me is how precisely the patent's mechanisms map to practitioner behaviors I was already recommending — building links with author names, pointing to About pages, and connecting entities to the businesses they represent. I was following this patent unknowingly.
I'll be explicit throughout about where patent evidence ends and practitioner inference begins.
The Honest Hedge
Every analysis has a threshold where certainty ends and inference begins. Here's where that line falls for this patent:
- Google designed a trust system where trust flows between entities (people, organizations, experts) — not between pages. Trust is computed as an eigenvector on the entity trust graph.
- The patent describes explicit trust mechanisms: trust buttons, trust lists, vanity lists, and trust inferred from visit patterns, email contacts, and chat lists.
- Trust is topic-specific: separate trust matrices per topic, with hierarchy aggregation. You can trust someone for medicine but not sports.
- Trust decays over time if not reaffirmed by the trusting entity.
- The API leak reveals
AnchorTrustedInfo.trustedTarget(boolean — explicit trusted source list) andhomePageInfo(4-level trust classification: NOT_TRUSTED → PARTIALLY_TRUSTED → FULLY_TRUSTED). - The inventor is Ramanathan Guha — creator of RSS, former Netscape and Apple, architect of multiple Google knowledge systems.
- Filed May 9, 2006. Granted October 13, 2009. The parent patent remains active with all maintenance fees paid through Year 12.
- The patent's entity trust graph is one of the clearest early algorithmic expressions of the same problem-space later described publicly through E-E-A-T — Experience, Expertise, Authoritativeness, and Trustworthiness. The concepts map closely to the patent's trust mechanisms.
- The API attribute
authorityPromotioninCompressedQualitySignalsappears to be the downstream output of entity trust evaluation — the trust factor the patent describes being applied to base information retrieval scores. - The patent's topic-specific trust matrices are the entity-level equivalent of
siteRadiusand centralized topicality from the API leak — same concept, different layer. - Three of four continuation patents expired due to non-payment of fees, suggesting the specific mechanisms (trust buttons, annotation databases) were absorbed into the Knowledge Graph and structured data systems.
- Whether the eigenvector-based trust rank computation described in the patent is still used in its explicit mathematical form, or whether graph neural networks now learn entity trust relationships from data.
- How Google currently defines the "entity trust graph" — whether it's purely derived from links, or from entity co-occurrence in text, structured data, Knowledge Graph connections, or all of the above. (All of the above would logically produce the most reliable results — similar to how Google uses five independent content quality systems that cross-validate each other.)
- The exact weighting of entity trust within the Q* composite quality score.
- Whether topic-specific trust matrices still exist as separate computations or have been absorbed into general-purpose neural entity understanding.
- The degree to which Google's modern implementation relies on the annotation/label system described in the patent versus Schema.org structured data and Knowledge Graph entity tagging.
Patent Metadata
The SEO industry commonly refers to "TrustRank" as a single concept. In practice, there are at least two distinct trust-related ranking mechanisms in Google's patent portfolio: (1) this patent (US7603350B1), which describes trust propagation between entities, and (2) the Seed Distance patent (US9953049B1), filed four months later, which measures trust propagation through the link graph. Both were filed in 2006, by different inventors (Guha vs. Hajaj), for the same ranking system. The SEO industry's "TrustRank" concept typically conflates the two — and often attributes entity-level trust properties to what is actually a link-graph distance model. The term "TrustRank" itself was originally coined in a 2004 Yahoo! / Stanford research paper by Gyöngyi, Garcia-Molina, and Pedersen — describing a link-graph spam detection technique. A third concept entirely.
What This Patent Does (Plain English)
Most ranking patents start from pages and links. This one starts from people. The patent describes a system where individual entities — experts, authors, organizations — create labels for web content. Other entities express trust in those labelers. Google computes a trust rank for each entity, then uses those trust ranks to re-rank search results.
Here's what the system does:
- Entities label content — An expert creates an annotation associating a descriptive tag (like "professional review" or "symptoms") with a specific URL or URL pattern. These annotations are collected by crawling expert websites or through an annotation API.
- Users express trust — Users declare which experts they trust, through explicit mechanisms (clicking trust buttons on expert sites, maintaining trust lists) or through behavioral inference (visit frequency, email contacts, chat lists).
- Trust matrix computed — All trust relationships are stored in a square trust matrix M, where M[i][j] represents entity i's trust in entity j. Trust rank is computed as the eigenvector of this matrix — essentially, PageRank applied to people instead of pages.
- Trust ranks aggregated — When a user searches, the system retrieves documents matching the query, checks which entities labeled those documents with matching labels, and aggregates the trust ranks of those entities into a trust factor per document.
- Search results re-ranked — Each document's base information retrieval score is multiplied by its trust factor. Documents endorsed by highly trusted entities rise; documents endorsed by untrusted or unknown entities stay low.
The Entity Trust Graph: PageRank for People
The heart of this patent is a trust matrix — a square matrix M where every entity in the system has a row and a column. If entity i trusts entity j, M[i][j] = 1. If not, M[i][j] = 0. From this matrix, the system computes a trust rank for every entity using eigenvector decomposition — the same mathematical technique that powers PageRank. In Brin and Page's 1998 Stanford paper, PageRank is defined as the principal eigenvector of the normalized link matrix: pages are states in a Markov chain, links are transitions, and the steady-state probability of the random surfer visiting any given page is computed via iterative matrix multiplication (the Power Method). This patent applies that identical computation — but the nodes are entities (people and organizations), not URLs, and the edges are trust relationships rather than hyperlinks.
The patent specifies multiple ways trust relationships can be established:
| Trust Signal | How It Works | Type |
|---|---|---|
| Trust button | A user clicks a button on an expert's website (implemented as an iframe with cookie-based user identification) | Explicit |
| Trust list | An entity publishes a list of other entities they trust on their website. (This maps directly to how "top 10" listicles function today — and how LLMs now consume and re-serve those rankings as authoritative truth. Much of GEO in 2026 echoes SEO in 2010.) | Explicit |
| Vanity list | An entity publishes a list of entities who trust them | Explicit |
| Visit frequency | The system infers trust from how often a user visits an entity's website | Inferred |
| Email contacts | Entities in a user's email contact list receive implied trust | Inferred |
| Chat lists | Entities in a user's instant messaging buddy list receive implied trust | Inferred |
| Transitive trust | If A trusts B and B trusts C, the system can infer A→C trust with a reduced value | Computed |
And critically, trust decays over time. The patent states that if a trust relationship is not reaffirmed — for example, by revisiting an expert's site and clicking the trust button — the "strength of a trust relationship" can "decay over time." Trust isn't permanent. It must be maintained.
This patent was filed in 2006. The explicit trust mechanisms it describes — trust buttons, trust lists published on personal websites, contact list scanning — were the 2006-era approach to a problem that has since been solved differently. Google launched Google+ in 2011 as what was arguably an attempt at this patent's trust button mechanism at scale. Google+ failed and was shut down in 2019. But the concept survived: modern transformer-based language models like BERT and MUM make it plausible that Google can now infer entity trust relationships from text, without requiring anyone to click a button — reading About pages, Schema markup, published articles, and conference mentions to derive the trust graph that Guha's patent required explicit declarations for. Before LLMs, you could build E-A-T faster through links than through a Knowledge Panel. In 2026, both channels are viable — and the signals likely compound.
The Annotation System: How Entities Label Content
The trust graph alone doesn't affect search results. The bridge between entity trust and document ranking is the annotation system — a database mapping entities to the content they've labeled.
An annotation, as the patent defines it, contains three components:
The patent describes two ways annotations are gathered:
- Crawling expert websites — A web crawler examines expert sites, extracting the links and their anchor text. If an expert links to a third-party camera review with anchor text "professional review," the crawler generates an annotation mapping that label to that URL with the expert as the entity.
- Annotation API — Experts can submit annotation files directly, containing URL patterns and labels in a structured format. The annotation database "includes thousands, even millions of such annotations."
Labels can include numerical values — ratings or significance scores that indicate how strongly the entity endorses the content. And the relationship between labels and URL patterns is many-to-many: a given label can be applied to documents matching multiple URL patterns, and a single URL pattern can carry many different labels from different entities.
The annotation database anticipates two systems that SEO practitioners interact with daily: structured data (Schema.org markup) and the Knowledge Graph. The patent describes entities creating free-text annotations linking labels to URLs — the same direction of travel that Schema.org later standardized with a shared vocabulary. The annotation database prefigures the Knowledge Graph: both connect entity identity to document relevance through structured relationships. Understanding this architectural lineage helps explain why structured data signals carry weight — they feed systems built on the same premise this patent formalized in 2006.
I've been telling people for years that author attribution on content matters — and this patent shows exactly why. You could build the off-page entity graph for the person, but if that entity isn't attributed to your website, you're not going to get the benefit. The off-page entity signals and the on-page attribution have to meet. A lot of people miss this because they don't reveal who is behind the business. Fix the bridge: make sure the entities you're building off-site are also clearly connected on-site through About pages, author bios, Schema markup, and attribution.
Topic-Specific Trust: The Precursor to Topical Authority
The patent doesn't treat trust as a single number. Trust is scoped to topics.
The patent states: "a user can have different levels of trust in an entity for different topics. For example, a user may trust an entity with respect to politics and economics, but not with respect to sports and entertainment." This means the system computes separate trust matrices per topic, with aggregation across topical hierarchies.
This is the entity-level equivalent of what the 2024 API leak reveals as siteRadius and centralized topicality at the content level. The content on your website has a topical radius — how focused or broad your coverage is. This patent describes the identical concept applied to the entity behind the content.
Think of it in four layers: your content has a topical radius (how narrow is your coverage?). Your site has a topical focus score (how concentrated is your domain?). Your link profile has topical relevance (are the sites linking to you in the same topic?). And your entity — the person or organization behind the site — has topic-specific trust (does the world trust this person on this topic?). This patent is the fourth layer. When all four align, you get compounding topical authority. When they diverge — when you publish content wider than the scope your entity's expertise can support — you dilute the signal across all layers.
I can illustrate this with an extreme example. Imagine a Chinese-speaking gifted children therapist who works exclusively with children under eight in Oregon. That entity is so narrow that if someone Googles "specialized psychotherapy for gifted children under eight in Oregon, Chinese-speaking," that person's website is the one answer directly on the money. All other results — a Montessori school, a general tuition center, a doctor who happens to cover the subject — will rank lower because their entity trust on that specific topic is broader. Now imagine that same therapist starts publishing blog posts about cryptocurrency, relationship advice, and home renovation. Their topical trust as an entity doesn't expand — it dilutes. The same principle applies at every scale. Topical authority isn't just about content strategy; it's about entity strategy.
Trust-Adjusted Ranking: How Trust Changes Search Results
The patent describes a concrete pipeline for applying entity trust to search rankings. Here's the process in FIG. 4:
- User submits a query — optionally including labels (e.g.,
cancer label:symptoms). - The search engine retrieves documents relevant to the query terms using a base information retrieval model (the patent names PageRank as "one suitable model").
- For each retrieved document, the system checks the annotation database for matching labels.
- The trust ranks of all entities who provided matching labels are aggregated.
- The aggregated trust rank becomes the document's trust factor.
- The trust factor is applied to the base IR score:
adjusted_score = base_score × trust_factor. - Documents are re-ranked by adjusted scores.
The patent specifies four different aggregation functions for combining trust ranks:
| Aggregation Method | Behavior | Effect |
|---|---|---|
| Linear weighting | Sum trust ranks with fixed weights | More endorsements = proportionally higher trust factor |
| Asymptotic (log) | Sum the log of trust ranks | Diminishing returns — prevents one dominant entity from inflating scores |
| Decaying weight | Weight decreases with each additional label instance | First endorsement matters most; subsequent ones are progressively less impactful |
| Sigmoid | S-curve aggregation | Threshold effect — trust jumps sharply once a critical mass of endorsements is reached |
And here's a detail that surprised me: the patent explicitly states that even queries without labels can use trust-based ranking. "If certain annotations are applicable to a search result document, then the trust rank for the entities providing these annotations is retrieved, aggregated, and applied to the base information retrieval score of the document as well." The label system is optional. The trust factor applies to all searches.
The choice of aggregation function has massive implications. If Google uses linear weighting, volume of endorsements matters proportionally. If they use decaying weight, the first expert to vouch for your content matters far more than the tenth. If they use a sigmoid, there's a threshold — below it, trust signals barely register; above it, they compound sharply. In practice, what I see in ranking data looks most consistent with a sigmoid or decaying-weight model: sites with no entity trust signals struggle regardless of content quality, while sites that cross the trust threshold see disproportionate gains. But this is practitioner observation, not confirmed architecture.
Entity Trust SEO Implications: What This Means for Your Strategy
1. Build the Entity Graph, Not Just the Link Graph
When I was running automotive affiliate sites from 2016 to 2019, I built links with a shortened version of my name pointing to the homepage and About pages. Volume was minimal — under four links per month — but those sites performed well enough that I sold multiple of them for five figures each. I wasn't following a patent. I was following what worked: connecting a real person's name to a domain through off-page signals. This patent describes why that works. Entity-anchored links don't just pass PageRank — they populate the entity trust graph, telling Google's systems: "this named individual stands behind this content."
2. The Attribution Bridge Is Non-Negotiable
Building the entity graph off-site is only half the job. The entity also has to be attributed on-site. If your About page doesn't name the person behind the business, if your articles don't carry author bios with Schema markup, if the connection between the entity and the domain is unclear — you're making Google's job significantly harder. The system can likely still infer the connection, but you get less credit for it. The off-page entity signals and the on-page attribution have to meet cleanly. This sounds obvious, but I see clients miss it constantly. They invest in PR and link building without ever making clear on their own website who stands behind the content.
3. Entity Expertise Is Scoped, Not Universal
This patent's topic-specific trust matrices mean that authority isn't a single number. A medical doctor trusted as an entity for "pain management" doesn't automatically carry trust for "financial planning." The moment you start publishing content outside the scope your entity's expertise can support, you don't expand your entity trust — you dilute it. Board certification, university affiliations, published work, industry membership — these are all entity trust signals that compound within a specific topic and weaken when stretched too broadly.
4. Run the Three-Step Entity Trust Audit
Here's the first thing to check after reading this article:
- Go to Ahrefs — enter your website and look at how many links point to your site using your name (or your key entity's name) as anchor text.
- Go to Search Console — check how many queries bring traffic from your name or your brand entity.
- Go to your own website — see exactly how your name is attributed, where it appears, and how well it connects to E-E-A-T signals (About page, author bios, Schema markup, credentials).
If there's a gap between your off-page entity presence and your on-site attribution, you've found one of the most impactful things to fix.
5. The GML Case Study — In Real Time
I'm not offering a polished retrospective here. This is a "physician, heal thyself" (Luke 4:23) moment. GetMeLinks got hit in the December 2025 core update. Our content was thin, overly commercial, and there was little attribution to the actual people, systems, and methodology behind it — even though all of those things existed. We did a very poor job of displaying it. The personal site you're reading right now, and the agency site rebuild, are both actively in progress to close exactly the entity attribution gap this patent predicts would matter. This is the case study being written in real time.
US9953049B1 (Seed Distance) — If this patent measures trust between entities, Seed Distance measures trust through the link graph. Same concept, different dimension. Both were filed in 2006, four months apart, by different Google engineers.
US8577893B1 (Reference Contexts) — Reference Contexts evaluates the editorial context around links — the quality of how entities embed their endorsements. The contextual layer on top of trust.
US7346839B2 (Historical Data) — Trust changes over time. Historical Data monitors temporal anomalies that could indicate trust manipulation. This patent's trust decay mechanism is the entity-level complement to Historical Data's temporal analysis.
Google API Leak Cross-Reference
The 2024 Google API leak — first reported by Rand Fishkin and investigated by Mike King at iPullRank — reveals multiple attributes that align with this patent's entity trust mechanisms:
| Patent Mechanism | API Attribute | Alignment |
|---|---|---|
| Trusted entity list (seed entities) | AnchorTrustedInfo.trustedTarget (boolean) | ✅ DIRECT TEXTUAL MATCH — explicit boolean trusted-source flag |
| Trust score from entity graph | AnchorTrustedInfo.trustedScore (float) | 🔶 STRUCTURAL ANALOGUE — fraction of pages with "newsy" anchors; shares the shape of aggregated trust rank |
| Trust validation | AnchorTrustedInfo.matchedScore (float) | 🔶 STRUCTURAL ANALOGUE — KL-divergence between spam and non-spam anchor distributions |
| Trust measurement count | AnchorTrustedInfo.trustedMatching (integer) | 🔶 STRUCTURAL ANALOGUE — count of matching trusted sources |
| Trust decay / demotion | AnchorTrustedInfo.trustedDemoted (integer) | 🔶 STRUCTURAL ANALOGUE — demoted trust signals; architecturally rhymes with trust decay |
| Homepage trust level | homePageInfo (enum: 0–3) | ✅ DIRECT TEXTUAL MATCH — 4-level classification: NOT_HOMEPAGE → NOT_TRUSTED → PARTIALLY_TRUSTED → FULLY_TRUSTED |
| Trust factor → base score | CompressedQualitySignals.authorityPromotion | 🔷 MODERN SYSTEM EXTENSION — Q* composite boost from trust signals |
| Anti-trust (independence) | AnchorsAnchorSource.indyrank (uint16) | 🔷 MODERN SYSTEM EXTENSION — independence score; defense against trust collusion |
✅ DIRECT TEXTUAL MATCH = the API attribute name directly corresponds to the patent mechanism. 🔶 STRUCTURAL ANALOGUE = the API attribute shares the same shape and purpose but may serve a different or broader system. 🔷 MODERN SYSTEM EXTENSION = the API reveals a signal not described in the patent but architecturally consistent with it.
The homePageInfo attribute reveals a four-level trust classification for homepage entities: NOT_HOMEPAGE → NOT_TRUSTED → PARTIALLY_TRUSTED → FULLY_TRUSTED. This is the closest structural analogue in the API leak to the patent's trust rank output — a categorical signal that shares the same shape as entity-level trust classification. The jump from NOT_TRUSTED to PARTIALLY_TRUSTED, and from PARTIALLY_TRUSTED to FULLY_TRUSTED, may correspond to the kind of observable ranking thresholds that practitioners see in site performance data. Crossing from "untrusted" to "partially trusted" would be the entity trust equivalent of getting your first seed-adjacent link in the Seed Distance patent — the entry ticket to being scored.
Citation Network
Patent Family Chain
US7603350B1 (this patent, filed 2006, granted 2009 — active) → US8352467B1 (2013, expired) → US8818995B1 (2014, expired) → US10268641B1 (2019, expired)
Google let three of four continuation patents lapse by not paying maintenance fees — while keeping the original parent active. This is unusual. The most likely explanation: the specific refinements in the continuations (trust button implementations, annotation database optimizations) were absorbed into the Knowledge Graph and structured data systems, making those particular claims unnecessary to maintain. The parent patent's broader claims about entity trust propagation and eigenvector-based trust ranking remain active — and those broader concepts remain architecturally relevant.
Key Patent Citations (Cited by This Patent)
| Patent | Relevance |
|---|---|
| US6285999B1 | The original PageRank patent — cited as "one suitable information retrieval model" for base ranking |
| US6636854B2 | Modifying search result ranking based on user-specified criteria |
| US7031961B2 | Method for ranking web page search results by criteria |
Related Articles on This Site
- US9953049B1 (Seed Distance) — Seed Distance measures trust through the link graph: how far is your page from sites Google already trusts? This patent measures trust at the entity level: how much do the people who endorse your content get trusted by others? They're the link-graph and entity-graph dimensions of the same trust architecture, filed four months apart.
- US8577893B1 (Reference Contexts) — Reference Contexts evaluates the editorial quality of how entities embed their endorsements. If entity trust is the who, Reference Contexts is the how — the contextual layer that validates whether the endorsement is genuine editorial support or manufactured placement.
- US7346839B2 (Historical Data) — Trust changes over time. This patent's trust decay mechanism has a temporal complement in Historical Data: if trust relationships appear or disappear suddenly, temporal analysis may flag the anomaly.
- US7716225B1 (Reasonable Surfer) — The Reasonable Surfer determines how much equity each individual link passes based on ML-learned per-link weights. Entity trust is the social layer (who endorses this content?); Reasonable Surfer is the structural layer (how much equity does each link carry?). A link from a high-entity-trust author in an editorial placement with high wᵢ is the trifecta.
- US9268820B2 (Knowledge Panel) — Entity Trust determines which sources Google treats as authoritative. The Knowledge Panel patent describes how that recognition gets displayed — the presentation layer that assembles structured entity cards alongside search results. A high-trust reference page is more likely to become a
cdocsource for the KP. - US8682892B1 (Implied Links) — Entity Trust determines who endorses your content. Implied Links determines whether the endorsement needed to be a hyperlink. Trust is the quality check on the endorser; implied links are the detection channel for the endorsement — unlinked brand mentions count alongside backlinks.
- NavBoost Deep Dive — Entity trust signals feed into ranking, but NavBoost provides the behavioural validation layer. A page can have high entity trust scores, but if users consistently bounce — if the engagement data contradicts the trust signal — NavBoost re-ranks accordingly. Trust gets you in the door; NavBoost determines whether you stay.
Entity Trust: What Doesn't Matter as Much as SEOs Think
The nature of this patent is simple: people trust people who are experts on the precise problem they themselves are facing, and the more times those experts help them solve it, the more trustable they become. Google took that fundamentally human reality and made it math. The patent's eigenvector computation, the trust matrices, the annotation databases — those are all flavor. The nature hasn't changed in 20 years, and it won't change in the next 20.
And this patent doesn't diminish the importance of links. Links remain a critical trust signal — the Seed Distance patent filed four months later proves Google was building both dimensions simultaneously. What this patent adds is the entity layer on top of the link graph: not instead of links, but in addition to them. The question isn't "links or entities?" — it's "do the entities behind your links carry trust in the topic you're trying to rank for?"
The flavor — trust buttons, explicit trust lists, free-text annotations — that was the 2006 approach. Google tried to build the explicit trust mechanism with Google+ in 2011 and shut it down in 2019 when it failed. But they didn't need Google+ to succeed. They just needed BERT. Once transformers landed, Google could infer entity trust from language rather than requiring people to click buttons and declare who they trusted. The destination was always right. The vehicle changed.
Here's what struck me: this patent was ahead of the technology by nearly two decades. Between 2006 and roughly 2023, Google couldn't fully do what Guha described at the entity level. When the August 2018 core update hit — nicknamed "Medic" by Barry Schwartz because of its outsized impact on health and medical sites — Marie Haynes became the key analyst connecting it to E-A-T and the Quality Rater Guidelines. Charles Floate pushed back, arguing that Google still used backlinks to measure E-A-T because they couldn't compute it from entity signals alone — and at the time, he was right. Before LLMs, you could build E-A-T faster through links than through a Knowledge Panel. Only now, with the combination of neural language understanding and the Knowledge Graph, is entity-level trust computation becoming viable at scale. And even now, links remain a signal layer.
The SEO industry chases Domain Rating as if it's the signal. It isn't. DR is a third-party proxy metric — it measures aggregate link power, and it tells you nothing about who stands behind the content, what topics they're trusted for, or whether Google sees a coherent entity connecting the off-page signals to the on-site attribution. Two DR 60 sites can have wildly different entity trust profiles. One has a named author with published work, Schema markup, and Knowledge Graph presence. The other is a faceless content farm. This patent explains the structural difference between them.
The reality of 2026 SEO was shaped 20 years ago. If a system that outlines the thinking behind entity-level trust was already formalized in a Google patent in 2006, consider: what patents has Google filed in the last five years? What thinking are they formalizing right now that will define the next decade? There are no big pivots coming. There never were. The path just gets narrower on what works, and wider on what doesn't — by gathering more signals, not fewer.
Frequently Asked Questions
What does patent US7603350B1 actually do?
It describes a search ranking system where trust flows between people (entities), not between pages. Experts label web content with descriptive tags, users express trust in those experts, and the system computes a trust rank for each entity using eigenvector decomposition on the trust matrix. Documents endorsed by highly trusted entities receive boosted ranking scores.
Is this the same as "TrustRank"?
Not exactly. The SEO industry's "TrustRank" concept conflates two distinct Google patents: this one (entity-level trust between people) and the Seed Distance patent (US9953049B1) (link-graph trust measuring proximity to seed sites). Both were filed in 2006, by different inventors. The term "TrustRank" was originally coined in a 2004 Yahoo! / Stanford research paper by Gyöngyi, Garcia-Molina, and Pedersen, describing link-graph spam detection — a third concept entirely. When someone says "TrustRank," clarify which mechanism they mean.
Is this patent the algorithmic basis for E-E-A-T?
It appears to be the closest algorithmic precursor to E-E-A-T. The patent's mechanisms map directly: entity expertise (via topic-specific trust), experience (via trust lists and visit patterns), authoritativeness (via eigenvector-based trust rank), and trustworthiness (via explicit trust relationships that decay over time). E-E-A-T as defined in Google's Quality Rater Guidelines was formalized in 2014 — eight years after this patent was filed. The patent describes the algorithmic skeleton; E-E-A-T puts the human-evaluation flesh on it.
How do I build entity trust for my website?
Build at three levels simultaneously. Off-site: get links and mentions that use your name (or your key entity's name) as anchor text, pointing to your homepage and About pages. On-site: clearly attribute the entity on your website — About pages, author bios, Schema markup (Person, Organization), credentials, and topical depth. Knowledge Graph: claim and maintain your Google Business Profile, earn Wikipedia references, publish in indexed industry sources. The off-page and on-page signals must meet — building one without the other leaves the bridge incomplete.
Why did Google let three continuation patents expire?
Three of four continuations (US8352467B1, US8818995B1, US10268641B1) expired due to non-payment of maintenance fees. The most likely explanation is that the specific mechanisms described in those continuations — trust button implementations, annotation database optimizations — were absorbed into the Knowledge Graph and structured data infrastructure, making the specific claims unnecessary to maintain. The parent patent's broader claims about entity trust propagation remain active.
Does entity trust matter more than links?
They're different dimensions of the same system. Links measure structural trust through the page graph (Seed Distance). Entity trust measures social trust through the entity graph (this patent). Both contribute to the Q* composite quality score. Before 2023, links were the faster and more concrete trust signal. With modern LLMs, Google can now infer entity trust from language — making both channels viable. The most structurally resilient strategy builds both simultaneously.
What's the "attribution bridge" and why does it matter?
The attribution bridge is the connection between your off-site entity signals (links with your name, media mentions, Knowledge Graph data) and your on-site entity attribution (About pages, author bios, Schema markup). If you build a strong entity graph off-site but don't attribute that entity on your own website, you're making Google's job significantly harder. The entity trust factor can't be applied to your documents if the system doesn't know which entity stands behind them. Many businesses invest heavily in PR and link building while neglecting on-site attribution — leaving the bridge incomplete.
