Last Updated on April 21, 2023
Discover how Google could assess the credibility of content by considering the author’s expertise, authority, trustworthiness, and experience.
In this article, we delve into the possible ways that Google might use authors’ experience, expertise, authoritativeness, and trustworthiness (E-E-A-T) to evaluate content.
Understanding Google’s E-E-A-T
Google has emphasized the importance of the E-E-A-T concept in enhancing search result quality and the user experience within the search engine results pages (SERPs). Various on-page factors, including content quality, link signals such as PageRank and anchor texts, and entity-level signals, all contribute significantly.
Unlike document scoring, the E-E-A-T concept doesn’t concentrate on evaluating individual content pieces. Instead, it has thematic relevance connected to the domain and the creator entity, irrespective of search intent or specific content.
Ultimately, E-E-A-T is an independent influence factor that operates regardless of search queries. It mainly pertains to thematic areas and functions as an assessment layer that evaluates content collections and off-page signals related to entities like people, organizations, companies, and their domains.
Recognizing the Significance: Author as the Content Source
Prior to the emergence of E-E-A-T, Google attempted to integrate content source ratings into search rankings. For example, the Vince update of 2009 provided a ranking edge to content produced by brands. Google endeavored to gather signals for author ratings (e.g., via a social graph and user ratings) through initiatives like Google+, which is no longer operational.
Over the past two decades, multiple Google patents have made direct or indirect references to content platforms like Google+ and other social networks.
Assessing the source or author of a content piece based on E-E-A-T criteria is an essential measure to enhance the quality of search results. Given the prevalence of AI-generated content and traditional spam, it is illogical for Google to incorporate subpar content in the search index.
The greater the volume of content that Google indexes and processes during information retrieval, the more computing power it necessitates. E-E-A-T enables Google to rank at a wider entity, domain, and author level without the need to crawl each content piece.
At this higher-level perspective, content can be categorized based on the originating entity and assigned more or less a crawl budget. Google can also utilize this approach to exclude entire groups of content from indexing.
Methods that Google Can Use to Identify Authors and Assign Content Attribution
Authors are categorized as a type of person entity. It is important to differentiate between recognized entities that are already documented in the Knowledge Graph and unverified or unknown entities stored in a knowledge repository like the Knowledge Vault.
Google has the ability to extract entities from unstructured content using machine learning and language models, even if they are not yet included in the Knowledge Graph. This process, known as named entity recognition (NER), is a subtask of natural language processing.
Named entity recognition (NER) identifies entities in the text by analyzing language patterns and assigning them to specific entity types. Typically, nouns are considered as entities.
Current natural language processing systems utilize word embedding techniques such as Word2Vec for this purpose.
Each word in a text or paragraph is represented by a vector of numbers in modern information retrieval systems, using techniques like Word2Vec. Entities can be represented in the same way as node vectors or entity embeddings, using methods like Node2Vec or Entity2Vec.
Part-of-speech (POS) tagging assigns words to their respective grammatical classes such as nouns, verbs, prepositions, and others.
Typically, nouns are considered entities. In a sentence, subjects are the primary entities while objects are secondary entities. Verbs and prepositions serve as connectors to relate the entities to one another.
Entities and their relationships can be identified using natural language processing techniques.
By creating a semantic space, a better understanding and representation of the entity’s concept can be achieved.
The document embeddings are the counterpart to author embeddings, and they are compared with author vectors through vector space analysis, as explained in the Google patent titled “Generating vector representations of documents.”
Representing all types of content as vectors enable several functions, such as: comparing content vectors and author vectors in vector spaces, clustering documents based on their similarity, and assigning authors to their respective content.
The proximity between the document vectors and their corresponding author vector indicates the likelihood that the author authored the documents.
If the distance between the document vectors and the corresponding author vector is smaller than other vectors and meets a certain threshold, the document is attributed to the author.
By using this method, it is also possible to prevent the creation of a document under a false identity. After the attribution of the document vector to an author vector, the author entity can be assigned to the author vector, as previously explained, by utilizing the author name provided in the content.
Examples of key sources of information for authors:
- Pages on Wikipedia that cover individuals.
- Profiles attributed to authors.
- Profiles attributed to speakers.
- Social media profiles linked to individuals.
When you search for the name of a person entity on Google, the first 20 search results typically include Wikipedia articles, author profiles, and URLs of domains that are directly associated with the author.
On a mobile search engine results pages (SERPs), it is possible to observe the sources that Google directly associates with the person entity.
All search results listed above the social media profile icons are identified by Google as sources that have a direct association with the entity.
Google uses personal profiles on the web to contextualize authors and identify social media profiles and domains associated with them.
Author boxes or author collections present in author profiles assist Google in attributing content to authors. The author’s name alone is not adequate to identify them since it can lead to ambiguities. Ensuring consistency in author descriptions is important as it can help Google verify the legitimacy of the entity by comparing them to one another.
Google’s patents: Assessing Author Expertise, Authority, and Trustworthiness
The following patents provide insights into the potential approaches that Google may employ to recognize authors, attribute content to them, and assess their E-E-A-T:
Content Author Badges
This patent outlines a method for assigning content to authors through a badge system. An author badge is created for the content and linked to an identification (ID) such as the author’s name or email address. Verification of the author’s identity is performed through a browser add-on.
Read more about Content Author Badges per Google.
This patent was signed by Google in 2016 and has a term of up to 2036. However, as there have only been patented applications filed in the USA, it suggests that this methodology is not yet implemented in Google searches worldwide.
The patent outlines the representation of authors as vectors using training data.
The vector is composed of unique parameters that are identified based on the author’s typical writing style and choice of words.
By representing authors as unique vectors based on their typical writing style and choice of words, content that was not previously attributed to the author can be assigned to them, and similar authors can be grouped into clusters.
One or more authors can have their content ranking adjusted based on the user’s past behavior in search (such as on Discover) after the vector-based author identification has been performed.
As a rewrite: Content authored by previously discovered authors and authors with similar characteristics may receive a higher ranking in search results.
One of the key components of this patent is the use of embeddings, including both author and word embeddings.
Nowadays, embeddings have become the technological norm in the fields of deep learning and natural language processing. Hence, it is evident that Google would utilize such methods for author identification and attribution.
Read more about Author Vectors per Google.
Scoring Reputation of an Author
Google signed this patent in 2008, and it has a minimum term of 2029. The patent originally pertained to the now-defunct Google Knol project.
It is intriguing why Google refiled this patent in 2017 under the new title “Monetization of online content” after it was originally filed in 2008 and had a minimum term of 2029. The patent was initially related to the now-defunct Google Knol project, which was terminated by Google in 2012.
The patent pertains to the determination of a reputation score, which can consider the following factors:
- Author’s expertise level.
- Publication in prestigious media outlets.
- Quantity of publications.
- The freshness of recent releases.
- Author’s experience in an official capacity.
- Quantity of links generated by the author’s content.
An author may have multiple reputation scores for each topic and can have multiple aliases within a specific subject area.
Given that many of the concepts described in this patent relate to a closed platform like Knol, it is reasonable to conclude that the patent remains relevant to the current state of affairs.
Read more about the reputation of an author per Google.
The 2005 Google patent has a minimum term until 2026 and has been registered not only in the USA but also in Spain, Canada, and worldwide. This suggests that it could be applied in Google searches globally.
According to the patent, digital content is assigned to an agent (either a publisher or an author), and the content’s ranking is determined based on the agent rank, among other factors.
The agent rank is not influenced by the search query’s intent and is calculated based on the documents assigned to the agent and their backlinks.
Moreover, the agent rank is specific to a single search query, search query cluster, or entire subject area.
Read more about Agent Rank per Google.
The Reliability of an Online Content Author
The first signing of this Google patent was in 2008, and it has a minimum term of 2029. As of now, it has only been registered in the USA. Justin Lawyer developed it, and it is directly related to its use in searches. This patent also focuses on determining the Reputation Score of an author and has similarities to the aforementioned patent.
The patent outlines several factors that can be utilized to determine an author’s credibility algorithmically. It explains how search engines can rank documents based on an author’s credibility and reputation score. Authors may have different reputation scores for each topic they publish on.
The reputation score of an author is not dependent on the publisher. Similar to the previous patent, this patent also mentions links as a possible factor in E-E-A-T rating. The number of links to an author’s published content can impact their reputation score.
The patent suggests several potential indicators for a reputation score, including:
- The length of time the author has been creating content in a particular subject area.
- The author’s level of recognition.
- Ratings are given by users to the author’s published content.
- The quality of ratings received by the author’s content when published by other publishers.
- The volume of content created by the author.
- The recency of the author’s last publication.
- Ratings of the author’s previous publications on a similar topic.
Additional noteworthy information regarding the reputation score outlined in the patent includes the following:
- An author may have distinct reputation scores depending on the variety of topics on which they produce content.
- An author’s reputation score is not influenced by the publisher.
- The reputation score may be lowered if the content is duplicated or if excerpts are published multiple times.
- The quantity of links leading to the published content has the potential to affect the author’s reputation score.
The patent delves into the topic of author credibility and outlines several factors that can influence it:
- The author’s professional background and role within a credible company, if applicable.
- The relevance of the author’s occupation to the topics covered in their content.
- The level of education and training of the author.
- The amount of time the author has been publishing content on a topic.
- The number of articles the author has published on a given topic can suggest expertise.
- The recency of the author’s content, with more recent publications being favored.
- Any awards or recognition the author has received in their field.
Read more about the reliability of an online content author per Google.
The Re-Ranking of Ranked Search Results Using Systems and Methods
The patent, signed in 2013, has a minimum term until 2033 and has been registered worldwide, in addition to the USA, indicating that Google may use it. Chung Tin Kwok, who was involved in several E-E-A-T relevant Google patents, is one of the inventors.
The patent explains that search engines can consider an author’s contribution to a thematic document corpus, in addition to references to the author’s content, in determining an author’s score.
The patent outlines a method for re-ranking search results that take into account author scoring and citation scoring, where citation scoring is determined by the number of references to an author’s documents. Additionally, the author scoring considers the proportion of topic-related documents contributed by the author to a corpus.
The primary objective of the patent is to detect and penalize “copycats” by lowering their content in the search rankings. However, it can also be utilized for the overall assessment of authors.
Read more about this on Google.
Important Rating Metrics for Rating an Author
While the above patents outline several potential criteria for evaluating authors, there are other factors to consider as well. For example, the overall quality of an author’s content on a given topic can be a significant factor for E-E-A-T, as measured by user signals, links, and other quality signals at the content level. Additionally, factors such as PageRank or references to the author’s content, co-occurrences of the author in different types of content (such as podcasts, videos, and PDFs), and co-occurrences of the author in searches with relevant topics or terms may all play a role in author evaluation.
If you need help assessing your website’s search presence, schedule a FREE consultation with our team today. Or, get a FREE SEO Overview Report of your site.