Authority is computed per author and per topic by multiplying authorship percentage by topic weight on every document the author contributed to. The mathematical quantification of expertise that underwrites the E-A-T author signal.
Patent Overview
- Inventor
- Michael Jeffrey Procopio
- Assignee
- Google LLC
- Filed
- 2012-01-31
- Granted
- June 4, 2013
The Challenge
The Challenge
Author reputation has historically been treated as a single global score or ignored entirely. The challenge: quantify expertise per author and per topic at the same time, account for partial contributions on co-authored documents, and produce a signature that lets the ranker reward genuine subject-matter authority instead of generic name recognition.
- Author Reputation Is Treated As Generic — Per author, traditional reputation models produce a single global score that does not distinguish subjects the author actually knows from subjects the author has never written about.
- Co-Authorship Dilutes Signal — Per document, multiple contributors share the byline at different depths of involvement, but flat author credit treats every named contributor as equally responsible.
- Topic Drift Within A Document — Per document, a single page often covers several topics at different weights, so attributing the entire document to one topic loses fidelity.
- Cross-Document Aggregation Is Missing — Per author, a meaningful authority score depends on aggregating contributions across every document the author has worked on, not on judging one document at a time.
- Expertise Is Not A Boolean — Per (author, topic) pair, expertise lives on a continuous scale that traditional models cannot represent because they lack the multiplicative authorship-by-topic structure.
Innovation
How The System Works
The system identifies authors and authorship percentages per document, extracts per-document topic weights, multiplies authorship percentage by topic weight to produce per (author, topic, document) contributions, and aggregates those contributions across every document the author touched to form a per-author per-topic authority signature.
- Identify Authors Per Document — Per document, named contributors are resolved to canonical author identities across the corpus.
- Assign Authorship Percentages — Per (document, author) pair, an authorship percentage encodes how much of the document the author is responsible for.
- Extract Topic Weights — Per document, topic weights record how strongly the document is about each topic in the system's topic space.
- Compute Per-Document Contribution — Per (author, topic, document) triple, contribution equals authorship percentage multiplied by topic weight.
- Aggregate Across Documents — Per (author, topic) pair, contributions are summed across every document the author worked on to form a topic-authority score.
- Store Author Topic Signatures — Per author, a vector of topic-authority scores forms the author's full expertise signature across the topic space.
- Feed Signatures Into Ranking — Per query, the topic-authority of each candidate document's contributors becomes a signal blended with content and link inputs.
Expertise Is A Number, Computed Per Author And Per Topic
The patent's load-bearing idea is that author authority is not a single trait. It is a per-topic signature built by multiplying the depth of an author's contribution to each document by how much that document is about each topic, then aggregating across every document the author touched.
Authorship Percentage Times Topic Weight, Aggregated
Per (author, topic) pair, authority equals the sum across documents of authorship percentage multiplied by topic weight. The arithmetic is simple. The consequence is that expertise becomes a measurable scalar the ranker can read.
- Authorship Percentage — Per (document, author) pair, how much of the document the author is responsible for.
- Topic Weight — Per document, how strongly the document is about each topic.
- Aggregated Signature — Per author, contributions sum across all documents to form a per-topic authority vector.
Technical Foundation
Technical Foundation
The patent specifies author identity resolution, authorship-percentage assignment, topic-weight extraction, multiplicative per-document contribution, cross-document aggregation, and downstream use of the resulting signature in ranking.
- Author Identity Resolution — Per byline, named contributors are resolved to canonical author identities across the corpus so contributions can aggregate correctly.
- Authorship Percentage Model — Per (document, author) pair, a numeric percentage reflects the author's depth of involvement, distinguishing lead authors from supporting contributors.
- Topic Weight Extraction — Per document, topic weights are derived from content analysis and represent the document's strength on each topic.
- Multiplicative Contribution — Per (author, topic, document) triple, the contribution score equals authorship percentage multiplied by topic weight.
- Cross-Document Aggregation — Per (author, topic) pair, contributions are summed across every document the author touched to form the authority signature.
- Ranking Signal Integration — Per (query, document) pair, the per-author per-topic authority of contributors is read by the ranker as an authority signal.
The Process
The Process
From a corpus of documents with authorship and topic metadata, the system computes per (author, topic, document) contributions, aggregates them into per-author per-topic signatures, and exposes the signatures to the ranker at query time.
- Ingest Document Corpus — Per document, bylines, content, and metadata enter the pipeline.
- Resolve Author Identities — Per byline, contributors are mapped to canonical author records.
- Assign Authorship Percentages — Per (document, author) pair, an authorship percentage is recorded.
- Compute Topic Weights — Per document, topic weights are derived from content.
- Compute Per-Document Contributions — Per (author, topic, document) triple, authorship percentage is multiplied by topic weight.
- Aggregate Author Signatures — Per (author, topic) pair, contributions sum across the author's full document history.
- Expose Signatures To Ranker — Per query, the ranker reads per-author per-topic authority for the contributors of each candidate document.
Quality Control
Quality Control
Author-level authority scoring introduces identity, gaming, and cold-start risks. The patent specifies safeguards to keep the signature honest.
- Author Identity Verification — Per author record, identity resolution is checked against signals including consistent bylines, profile links, and structured author markup before contributions are aggregated.
- Minimum Contribution Threshold — Per (author, topic) pair, a signature dimension is used only after enough contribution has accumulated for the score to be statistically meaningful.
- Topic Weight Sanity — Per document, topic weights are bounded and normalized so a single document cannot inject extreme authority for an author.
- Authorship Percentage Bounds — Per document, authorship percentages across contributors sum to a fixed total so credit cannot be inflated by claiming overlapping shares.
- Cross-Topic Sanity — Per author, signatures are cross-checked across topics so suspicious concentration in a manipulated topic can be flagged against the author's broader history.
Real-World Application
Per-author per-topic authority signatures are the mechanical foundation for treating expertise as a measurable input to ranking. The same author can rank as a strong authority for one topic and a non-authority for another, and the ranker reads both states from the same signature vector.
- Per (author, topic) Signature Granularity — Authority is computed for every author across every topic in the topic space.
- Multiplicative Contribution Formula — Authorship percentage multiplied by topic weight, summed across documents.
- Vector Author Signature — Each author carries a per-topic authority vector exposed to the ranker.
Why Topic-Specific Authority Beats Generic Reputation
Per (author, topic) pair, the signature captures whether an author has actually contributed to documents about that topic at meaningful depth. Generic reputation models cannot tell a topic specialist apart from a general writer, so they reward visibility instead of expertise.
Why Recurring Contribution Compounds
Per author, each new on-topic document with substantive authorship percentage adds to the topic dimension of the signature. Single-shot contributions register once. Recurring contributors accumulate authority that single guest appearances cannot match.
<\/section>What This Means for SEO
What This Means for SEO
Per-author per-topic authority signatures mean expertise is quantified by the ranker at the author level, not just at the page or site level. Author strategy must be planned around specific topics with depth, recurring contribution, and clear attribution.
- Author Authority Is Topic-Specific — An author can be a strong authority on one topic and unknown on another inside the same signature vector. Plan the author's contribution path around the topics where authority should accumulate instead of treating reputation as a single global trait.
- This Is The Structural Basis For E-A-T Expertise — Author signals matter to the ranker, but in a topic-specific way that mirrors how the E-A-T expertise dimension is described publicly. Generic author bios that do not align with a clear topic produce a thinner signature than focused, on-topic bylines.
- Authorship Percentage Matters — A sixty-percent contributor signals deeper involvement than a ten-percent contributor on the same document. Author credits that obscure the lead author, hide co-authorship structure, or list everyone equally dilute the signal the ranker can read.
- Topic Weighting Penalizes Tangential Pieces — Per document, topic weight controls how much each contribution feeds the per-author signature for that topic. Writing tangential pieces that mention the topic in passing builds little authority. Focused, on-topic contributions concentrate the per-topic signature.
- Recurring Contributors Compound — Single-shot guest posts register one document's worth of contribution. The author's full signature builds across many on-topic pieces, so a recurring contribution cadence on a single topic outperforms scattered one-off appearances across many sites.
- Cross-Platform Identity Aggregates The Signature — Per author, contributions aggregate across sites when the system can identify the author across them. Consistent byline strings, linked author profiles, and structured Person markup help the system attribute contributions to one canonical author record.
- This Lineage Feeds Agent Rank And Site Topical Authority — Per-author per-topic authority connects to the earlier Agent Rank author-reputation patent and to site-level topical authority patents authored elsewhere at Google, layering site-topic authority, author-topic authority, and global link authority into a single ranking stack. Earning recognition at the author layer compounds with the site layer and the link layer rather than substituting for them.