Author reputation in SEO: the Google patent that ranks content by the credibility of its author

Author reputation is a score given to an author from the reviews that other people give to the author's content, used to rank that content in search results. Patent US 8,150,842 B2, "Reputation of an author of online content," describes it. Google filed it on December 11, 2008, from a provisional application of December 12, 2007, and the patent was granted on April 3, 2012.

What does patent US 8,150,842 describe?

Patent US 8,150,842 describes a content platform that scores each author, verifies who the author is, and ranks the author's content with both. Its 4 inventors are William C. Brougher, Nathan Stoll, Sepandar D. Kamvar and Michael D. Dixon. The patent holds 51 claims and 19 drawing sheets.

The abstract defines the core: for each content item, "a reputation score is determined for the author," based "at least in part on one or more reviews of the online content item provided by one or more reviewers other than the author." In response to a query, "a ranking of the online content item in the set is determined based at least in part on the reputation score of the author."

The claims go further than the abstract. In the 3 independent claims (1, 18 and 35), the reputation score rests on 2 inputs: the scores that reviewers other than the author give to the author's content, and an authentication score, "a function of determinations made to identify that the author is who the author purports to be." All 51 claims cover the reputation score. The credibility factor and the identity checks described below come from the patent's description, not from its claims.

What is Knol, and why does this patent describe it?

Knol was a Google service for signed articles written by experts, announced on December 13, 2007, opened in beta in July 2008 and closed in 2012. A "knol" (for "a unit of knowledge") was an article on a topic, written by an identified author, rated and reviewed by readers. The patent describes it in detail, and its drawings show a knol titled "Insomnia: Risk factors, diagnosis and treatment," by Rachel Manber, presented as director of the Stanford Sleep Disorders Center, with 3,476 views, 12 reviews and 117 ratings.

The patent explains the problem it addresses: "Anonymity on the internet is easy," but it "makes it easy for any person of questionable background to pass themselves off as knowledgeable," while making it difficult "for those who are actually knowledgeable to provide proof of identity and credentials."

Which factors build an author's reputation score?

The patent's summary and dependent claims list the factors that can build or adjust the reputation score:

1. Reviews by other people of the author's content.
2. The reputation of the reviewers: if reviewers have a reputation score, the author's score is based in part on it.
3. The fame of the author, which the description estimates from mentions of the author in a web index or in publications, the appearance of the author's name in query logs, or the number of links to the author's blog.
4. The publisher: the score is elevated when the content is published by a publisher "publishing only online content given a review exceeding a predetermined threshold."
5. The number of other content items published by the author.
6. The recency of the author's other content.
7. Previous reputation scores of the author, on other content.
8. The topic: previous scores on the same topic count, and an author can have "more than one reputation score" for different topics, with a different alias per topic, the aliases being related to each other.
9. Sub-scores: the score can hold "two or more sub-scores," each for a different quality of the content.
10. Plagiarism: the score "can be reduced if the author is determined to have included plagiarized content."
11. Authentication: in the independent claims, an authentication score is part of the reputation score. The identity can also be authenticated before any score is computed, and the score can grow with "the length of time an author has been an authenticated author."
12. Links: "the number of links to the online content item from other online content."
13. The web page score: claims 17, 34 and 51 base the score on "a web page score attributed to a web page on which the author's online content item is published."

The score "can be portable from one online publisher to another": a reputation earned with another publisher can be imported, at least where the author is properly authenticated.

How does the patent weigh reviews and activity?

The patent weighs reviews by the reputation of the people who give them, in "a PAGERANK-like technique applied to author reputations." Its example: if Stephen King gives 5 stars to another author, that author's reputation "will increase more than it would if an unknown with a small reputation did the same." The description adds 5 rules:

1. Ratings are cleaned. A user's ratings can be normalized when the user is too "nice" or too "mean," and a user whose ratings suddenly disagree with everyone else's can be suspected of trying to push a score up or down.
2. Unauthenticated users can be barred from ranking, commenting on or editing content, to prevent dishonest rankings.
3. Time and activity count. A user whose contributions average 3.0 over hundreds of contributions and many years "may have a substantially higher reputation score" than a user with the same 3.0 average from a handful of readers on a single article.
4. Original content weighs more than derivative content. Reviews and comments earn reputation too, but "normal authorship may be weighted more highly."
5. Some qualities are scored per topic, others across topics. A nuclear physicist can have a very high reputation on relativity and a very low one on home theatre systems, while a skill such as technical writing can be judged across all her topics.

What else does the reputation score do?

The reputation score also decides how much an author earns and how the author is displayed. In one aspect described in the summary, the author's share of advertising revenue on a content item is based at least in part on the reputation score: an author with no rating "may have a minor share," while authors "with a platinum or other high rating may receive the maximum share."

The description shows reputation as seals: no icon for new authors, silver, gold or platinum seals for experienced ones, and black seals for authors who receive bad ratings. Reputation can also score web pages in search results "along with other traditional factors such as PAGERANK scores, click-through rates, ratio of good scores, and spam scores."

For new content, the patent notes that a page nobody links to yet can score low "even if it is the best page in the world." Content flagged as authoritative on a topic can be crawled more often, and its search scores can "depend more on ratings from users than on ratings from other pages in the form of incoming back links."

What is the credibility factor?

The credibility factor is a score based on information about the author "verified to be true." It ranks content as well, it is tied to a specific content item, and it depends on the relevance of the author's background to the topic. The patent lists these kinds of verified information:

1. A reputation score and an authentication score for the author.
2. Membership in an organization, and information about the organization. An organization that requires a license (to practice law or medicine, for example) weighs more.
3. Employment, and its relevance to the content.
4. Education or training in a field, and its relevance. The accreditation and rank of the institution count too.
5. Other relevant publications by the author.
6. Citations of the author's content in publications by other authors, especially by authenticated authors or authors with high credibility factors.
7. Awards and recognition in a field.
8. Feedback from organizations about the author or the content.
9. Revenue information about the author's content, such as best seller lists or revenue generated.

The description adds the length of time as an authenticated author, the credibility of co-authors, and publication by a publisher that regularly publishes award-winning authors.

The patent's examples: a dentist can have a high credibility factor for an article on root canals, "but may have a lower credibility factor for an online content item written in the subject area of muscle cars." An author employed as a food inspector gains credibility on food safety inspection, while a mechanical engineer in the oil and gas industry gets "a neutral or negative influence" on the same topic. A symbol representing the credibility factor can be displayed next to the author and the content.

To verify the information, the author can answer a questionnaire and provide proof, such as a scanned copy of a degree, or a third party source can be contacted: organizations, educational institutions, telephone directories, or a crawl of the web for references to the author.

How does the patent verify an author's identity?

The patent verifies identity by checking the author's details against a third-party source. It describes 3 main checks:

1. A phone number: confirming with the telephone provider that the name and number match, and optionally giving the author a token, calling the number and requiring the person who answers to provide it. The description also mentions a reverse look-up, a comparison with the location of the user's IP address, and a token sent by text message or by mail.
2. A credit card: confirming with a credit agency that the name and card number are associated.
3. A tax identification number: confirming with a credit agency or the Internal Revenue Service.

Other signals help: "out of wallet" data that an impersonator is unlikely to know, an email address at an organization's domain, such as "nasa.gov" for a user who claims to work at NASA, or a long history in a community such as Google, eBay or Amazon. Prominent authors, such as reporters, can be contacted directly with a unique token, and an authentication the system initiates can earn a higher score than one the user initiates.

The authentication score can be a weighted sum of successful and failed attempts. Because some checks give false negatives, a user can earn a few points for repeated failed attempts with the same credentials. The score can rise when the user stays in good standing for a long time, and fall when other users, especially users with high reputations, question the user's authenticity. It can be shown as a badge, a number of stars or a rank on a leaderboard.

How does this patent relate to E-E-A-T?

The patent predates the vocabulary, but it describes the signals that experience, expertise, authoritativeness and trust refer to. Google's guidelines for its search quality raters ask raters to consider who created content and whether they have the experience or expertise for the topic. The patent shows one way a system can measure that: reviews by peers, verified credentials relevant to the topic, and citations.

The Agent Rank patent approaches the same goal with digital signatures and a graph of agents, and the trust patent ranks pages by the trusted entities that label them.

What does author reputation change for your SEO?

Author reputation means that content gains from an identified author whose expertise matches the topic and whom others review and cite. 7 consequences follow:

1. Publish under real, verifiable authors. In the claims, the reputation score includes an authentication score, and anonymous content gives the system no identity to authenticate.
2. Match authors to topics. Credibility depends on the relevance of the author's job, training and publications to the subject: assign each topic to the author qualified for it.
3. Show credentials on the page. Employment, education, publications and awards are the patent's verified information: state them in the author bio and in structured data, in a form that can be checked.
4. Earn reviews and citations from peers. Reviews from reputable reviewers and citations by other authors raise the score more than praise from unknowns.
5. Never publish copied content. Plagiarism reduces the author's reputation.
6. Build reputation topic by topic, over time. Scores can be kept per topic, and the length of time as an authenticated author and a long record of contributions count.
7. Publish where quality is high. The score can rise with the publisher's standards and with the score of the page that hosts the content.

An SEO audit from our team checks these 7 points for each author: byline, bio, credentials in structured data, and the match between the author's field and the topics assigned.

The patent describes what Google's system can do. It does not confirm that Google computes author reputation this way in its rankings today.