How does Google rate E-E-A-T?
Google uses E-E-A-T to gain a better understanding of a user’s search in order to better tailor results. We have 13 possible ways Google rates E-E-A-T on a website. In doing so, we divide the paths into three focus groups:
1. How Google uses data science and machine learning to evaluate E-E-A-T factors.
2. how Google uses links to evaluate the E-E-A-T factors.
3. how Google rates authors or publishers (websites) using the E-E-A-T factors.
Data Science and Machine Learning
1. quality of the content
Google uses E-E-A-T to rate the quality of the publisher, author or associated domain in relation to one or more topics. This is to evaluate the quality and relevance of the content. However, Google does not only evaluate the quality of the author or an editor with the help of E-E-A-T, but also the relevance of the content on the document level. For this purpose, classical methods of information retrieval such as text analysis and machine learning (ML) such as Rankbrain are used. However, the quality of content from different subject areas can influence each other both positively and negatively.
PageRank is an ML algorithm that helps Google determine the relevance of websites to a particular search query. Here, the algorithm assumes that a website that is linked to by many other high-quality websites is likely to contain high-quality content. A high number of backlinks from trusted websites can be taken as an indicator of authority within the E-E-A-T concept.
3. sentiment analysis of mentions and click-through rate.
Via Natural Language Processing, Google can perform sentiment analysis on the entity that results from a website or author. Thus, the author or website may be rated with more credibility when the mood is positive, and with less when the mood is negative.
Structured and unstructured data can be used as a source for sentiment analysis, such as reviews from Google Maps, TripAdvisor, Citysearch or Yelp.
Entities are stored in databases for sentiment analysis. There they are represented by tuples in the form of entity ID, entity types and one or more ratings. Different scores are assigned to the ratings, as well as additional information such as the author, which is calculated in the Ranking Analysis Engine.
The patent Sentiment Detection as a Ranking Signal for Reviewable Entities also discusses the use of interaction signals. They are intended to complement sentiment in terms of ranking as a factor, for example, by looking at user signals such as SERP CTR and dwell time.
4. trust based on knowledge
Google’s scientific paper on Knowledge-Based Trust deals with the credibility of websites determined by algorithms. It presents two methods for determining credibility: Analyzing links and checking for correctness of published information using data mining. According to Google, the latter new approach relies on endogenous signals – the correctness of factual information provided by the source.
Those sources that contain few false facts are considered trustworthy. In this way, Google wants to fill gaps in the evaluation of credibility, which were created by the fact that previously only links and user behavior were used. This is because less popular sources slip in the rating, although they provided correct and good information.
Using this approach, the trustworthiness of sources can be assessed independently of their popularity. Thus, websites that frequently provide incorrect information are devalued and those with correct information (i.e., those that are in the general consensus) are positively evaluated and rewarded. This way, Google tries to make websites that publish Fake News less visible.
5. trusted seed sites
If a publisher or an author is often linked on trusted see sites, it increases his authority in that topic – semantically related keywords are also included in this evaluation. In order for seed sites to be trusted, they must be manually and carefully selected and applied only in limited numbers. This is to avoid manipulations. The algorithm also takes into account the length of the link between the seed website and the document being evaluated. This can be determined by the position of the link, the degree of thematic deviation of the source and the number of outbound links of the source. It is interesting to note that websites that do not have a direct or indirect link to at least one seed website are not included in the evaluation at all.
6. anchor text from backlinks
According to Google, the anchor text of backlinks is not only a ranking signal for the target page. It is also used to fit into the theme of the domain. In the patent Search result ranking based on trust, Google refers to the use of anchor texts as an evaluation of trust. The patent describes how the ranking scoring of documents is supplemented with the help of a trust label. The information for the rating can come from the document itself or from documents that are referenced, for example through link texts. The trust labels are assigned to the URL and recorded in a database.
7. links on own website
Links to publications, interviews or similar on one’s own website also make it easier for Google to evaluate the E-E-A-T factors of authority and expertise or experience.
In the E-E-A-T concept, the author probably plays the leading role. Google also evaluates documents in terms of an author’s credibility and reputation. The reputation values of an author are composed of different points:
- Reputation values can arise from the different topics on which he publishes content. So it can have different reputations for different topics.
- The reputation rating of an author is independent of the publisher, e.g. of a particular website.
- Duplicate content and content excerpts published multiple times can downgrade an author’s reputation value.
- The reputation value of an author can be influenced by the number of links on published content.
Other signals that have an impact on the author’s reputation rating are:
- Duration of the track record of an author
- Name recognition of the author
- User rating of the published content
- Above-average content rating by another publisher
- Number of published contents
- Regularity of the contents
- Evaluation of topic-like content from the past
The credibility factor of the author should be evaluated with the help of information about the profession or role in the company. In addition, the author’s level of education also influences credibility.
The higher the name recognition, the more credible the author – and the higher his or her authority in a subject area. Google measures an author’s or publisher’s name recognition using algorithms that obtain data on the number of mentions and search volume for that name. In particular, local searches may be influenced by the name recognition of a publisher, such as a museum, hotel, or the like. In addition, the awareness of an author or publisher is measured from the information provided, for example, via links, articles or directories.
Mentions, for example in social networks, can also have an impact on the name recognition of an author or publisher – as long as they are of high quality, of course.
… with thematically relevant terms in videos, podcasts & documents
The frequency with which an entity appears in crawlable and interpretable content with terms from specific subject areas can help Google place an author or publisher in a thematic context.
These co-occurrences can be considered in the E-E-A-T evaluation by using the number of co-occurrences and the authority and trustworthiness of the sources. With the help of innovations like MUM, this content can be images, video, and audio, in addition to text content.
… with thematically relevant terms in search queries
Co-occurrences of entity and topically relevant terms in content can also help Google evaluate the E-E-A-T. Co-occurrences of search queries can also be an important signal here.
If many users search for “UnitedAds SEO consulting”, this may indicate that UnitedAds holds an authority for SEO consulting.
The Google patent Systems and Methods for Re-Ranking ranked Search Results describes a method by which search engines can include an author’s contribution to a thematic corpus of documents in the ranking. This refines search results according to an author scoring and citation scoring. Citation scoring is based on the number of references to an author’s documents and the proportion of content that an author has contributed to a corpus of documents.
Transparency is another signal Google uses to E-E-A-T an author or website. This transparency is provided, for example, through pages such as About Us, author profiles with images, physical addresses, or memberships.
The goal here is to show that there is a real organization or real people behind a website or author. In this way, expertise and experience are shown transparently and trust is established.
13. use of a https
Finally, one more way Google evaluates the E-E-A-T factors of a website or author, but it doesn’t quite fit into our three groups: Although https are otherwise only a minor ranking factor for Google, they are useful: Namely in terms of trust, i.e. the trustworthiness of a page. The more trustworthiness, the better.