27 Oct The Mechanics of E-A-T: How Google Patents Can Help Explain How E-A-T Works via @lilyraynyc
Ever since E-A-T has taken on a central role in the SEO discussion, it has become the source of many myths and misconceptions.
It has also proven to be a difficult topic for many in our industry to understand.
The reasons for this are clear.
Beyond sharing a select few resources about E-A-T, Google hasn’t explicitly confirmed which ranking factors are considered part of their E-A-T evaluations aside from PageRank and links.
It's almost like we look for signals that align with expertise, authoritativeness and trustworthiness. We should give that an acronym like E-A-T and maybe suggest people aim for this. Oh wait, we did: https://t.co/1fs2oIS54Lpic.twitter.com/xNL424dDdq
— Danny Sullivan (@dannysullivan) October 11, 2019
This creates a scenario where we’re left with many questions about the role that E-A-T plays in Google’s algorithms.
While Google has made it abundantly clear that its search quality raters do not directly influence Google’s search results, they haven’t answered many of the other pervasive questions surrounding E-A-T and the mechanics behind it, such as:
- How does Google identify authors and experts?
- How does Google determine if a website is classified as YMYL (Your Money, Your Life)?
- How do Google’s algorithms determine what is in line with scientific, medical, or historical consensus?
- Is E-A-T evaluated on the page, domain, entity, and/or company level?
Luckily, many of the answers to these questions can be gleaned by reading Google’s patents, which describe in detail the processes Google uses to rank pages (among many other things).
Several of Google’s patents – particularly those filed in recent years – contain information related to identifying authors, categorizing websites, and classifying levels of expertise, which can help explain how Google may be using E-A-T algorithmically.
A Disclaimer About Google Patents
It’s important to note that Google’s patents are also not an exact explanation of how their algorithms work.
Especially given that we don’t know precisely which patents they are actively using and for which Google products.
But, the patents can help to give us an understanding of what Google’s algorithms are capable of.
I reached out to Google patent expert, Bill Slawski, to explain how Google’s patents might be able to help us gain a deeper understanding of how E-A-T works.
We worked together to create this article.
Below are some of the most common questions about E-A-T and how Google’s patents can help to answer them.
1. How Does Google Know Who a Site’s Authors Are?
There are a variety of Google patents that help answer this question.
To start, Google’s filed the Agent Rank patent in 2007.
This patent, according to Slawski, can “potentially boost rankings for pages based upon the identity of authors or editors or commentators or reviewers on pages.”
The patent included the ability to identify these authors and experts (also known as “agents”) by their digital signatures, such as their byline, and rank content according to their combined reputation scores.
However, after the August 1, 2018 core update, Google’s John Mueller clarified that Google does not use individual author reputation as a ranking factor.
That said, Slawski notes that it’s important to distinguish here between “reputation” and “expertise” or “authoritativeness.”
Reputation is how others perceive an author.
Authority and expertise are attributes that Google itself evaluates and assigns to a given author or other entity.
While the concept of authorship has greatly evolved since that point, a look at Google’s recent patents indicates that they are still working on being able to identify authors online.
Google filed a patent in March 2020 called Generating Author Vectors.
It enables them to be able to identify authors throughout the internet based on their writing style alone, even when their names are not explicitly mentioned on the page.
In his article, Author Vectors: Google Knows Who Wrote Which Articles, Slawski explains that Google’s new patent uses a neural network system trained on a set of words that it uses to be able to identify an author, even without the text being labeled as being written by that author.
Google is then able to create an “author vector,” which can be used to:
- Characterize authors.
- Identify unique traits about their writing style.
- Identify other authors whose writing is similar.
If this patent is actively being used in Google’s organic search algorithms, this has a variety of interesting implications for SEO.
For example:
- Google may be able to identify if a page has falsely used an expert’s name as the author of an article they did not write
- Google may be able to assign authorship to articles where the author is not listed, based on the writing style alone
- Google may be analyzing the quality and writing style of a given expert to gauge what expert content should look like on that topic. Relatedly, Google has a patent that suggests that they can classify the quality of content using n-grams and comparing n-gram statistics between websites.
These implications suggest that faking good E-A-T is not as easy as it sounds.
Claiming your content was written or reviewed by experts is not sufficient to improve your E-A-T if the quality of the content itself does not actually reflect their expected level of expertise.
2. Does Google Only Look at On-Page Factors & Links to Identify Authors and Experts?
Interestingly, in early 2020, Google was granted a patent called Speaker Identification, which allows them to identify a speaker by using speech recognition.
Google does this by honing in on unique aspects of how that speaker communicates, such as their accent.
This patent is likely applied most in places like YouTube, where Google has an enormous database of audio and video content it can analyze to identify and obtain information about speakers.
Slawski points out this is part of a larger trend he has noticed during his 16 years analyzing patents:
“Google wants to index actual speakers and authors and websites treating each as an entity, understanding, and indexing each of those based upon the features which make them unique.”
This patent just goes to show how far Google’s focus on E-A-T can potentially go.
Text content in its organic search results is likely not the only source of information Google uses to evaluate experts.
The process can extend to audio, video, and possibly even images across all of Google’s products.
3. Can Google Only Recognize Entities Contained in Its Knowledge Graph?
We don’t exactly know how Google evaluates authors and other entities that are not included in its Knowledge Graph.
However, during Google’s recent video, Search on 2020, Google indicated that it would be using “data that’s spread across multiple sources” to help answer users’ questions.
Since 2018, Google has been working with the U.S. Census, the World Bank, and other data sources in an open database called Data Commons.
Google announced it will be taking things a step further by including this data as a “new layer of the Knowledge Graph.”
Google will be using natural language processing to better understand user intent and to map queries to relevant sources in Data Commons.
You can view a list of those data sets here in the Linked Open Data Cloud.
It’s hard to say how this will change Google’s Search Results, but given the vast amount of data available in Data Commons, this could have the potential for Google to recognize thousands of new entities not currently listed in its Knowledge Graph.
This is especially true for U.S. Census data, which contains information about millions of individuals Google likely doesn’t currently recognize as entities.
4. How Does Google Determine If a Person or Brand Is a Real Expert, or Authoritative in Their Field?
This is one of the most commonly asked questions about E-A-T, which gets at the heart of what many SEO professionals wonder about.
Is there a specific threshold one must meet to be considered a true expert or authority on a given topic?
As a reminder, there is no E-A-T score or YMYL score, which was specifically confirmed by Google last year.
But looking at Google’s recent patents can help us understand how Google might be working toward measuring the level of expertise of a given entity.
The patent that comes closest to answering this question is the Website Representation Vectors patent.
This, importantly, was filed at the same time as Google launched the now-infamous August 1, 2018 core update (informally labeled as “Medic”).
The patent tells us that Google is capable of classifying websites into various categories of expertise, such as expert, apprentice, and layperson, and rank pages based on the authoritativeness of the content found on those pages.
The patent claims that Google can use “any appropriate method” to generate these classifications, which is pretty open-ended.
But it also provides a few examples of how Google may do this, such as:
- Analyzing the text or images on the website.
- Looking at other website content (i.e., links).
- A combination of two or more of the above.
Another patent granted to Google in 2017, labeled Obtaining Authoritative Search Results, describes the process Google uses to rank authoritative sites for queries that require authoritative results, and how Google can distinguish authoritative sites from sites that are low-quality because they contain shallow content or too many ads.
Between the various patents listed above, it is clear that Google is at least working toward being able to identify authors (whether or not they are explicitly mentioned on the page), and evaluate their levels of expertise or authoritativeness on a given topic by analyzing a variety of on and off-page factors.
Google hasn’t explicitly confirmed that taking steps such as listing the author’s credentials in a biography or linking to the other places they have been mentioned online are factors they look at when measuring E-A-T.
However, Google indicated that “millions of tiny algorithms” work together to conceptualize E-A-T and YMYL.
So it’s fair to assume that these signals are part of that consideration process.
Mueller also recently confirmed that it’s important to look at what SEO pros are saying about E-A-T, when it comes to improving the performance of health and medical content.
At the very least, the commonly recommended approaches are good for users, but Mueller indicates that Google can pick up on them too.
More from @johnmu: Look at what SEOs are explaining about E-A-T. Understand how to best present your content, author profiles, & more. John can't guarantee that will improve rankings, but make sure you have all of those signals for users. Google can pick those up too: https://t.co/IPg9MI5r08
— Glenn Gabe (@glenngabe) October 16, 2020
5. How Does a Website Get Classified As YMYL?
The same patent listed above – Website Representation Vectors – is the best place to look when understanding what mechanisms Google uses to determine what content is considered YMYL: Your Money, Your Life.
In this patent, Google uses neural networks to understand patterns and features behind websites as a means of classifying those websites into categories such as health, medical, finance, and more.
As part of this process, Google can also determine the level of expertise required for certain topics, such as:
“[E]xperts in the knowledge domain, e.g., doctors, the second category of websites authored by apprentices in the knowledge domain, e.g., medical students, and a third category of websites authored by laypersons in the knowledge domain.”
While the patent doesn’t clearly delineate which categories are considered YMYL, it shows how Google categorizes sites into niches and evaluates authority accordingly.
The patent also indicates that, for some queries, Google may be limiting its retrieval of results to those which are included within a certain category of domains.
Here is Slawski’s explanation of how this might take place:
“If this process limits the number of sites that Google has to return search results from based upon which knowledge domain they might be in, it does mean that Google is searching through fewer sites to return results than Google’s entire index of the web…
The Search System may select, search, or both, data for only websites with a particular classification, reducing computer resources necessary to find search results, e.g., by not selecting, searching, or both, any website irrespective of classification.”
The above portion of this Google patent indicates that, for certain queries, Google may be looking within its established set of highly authoritative websites within a given category when it is ranking pages.
If you’ve ever wondered why you always see the same set of 10-20 authoritative domains on a given topic – especially for YMYL topics – this may be the reason.
6. Does Google Measure E-A-T on the Author, Page, Domain, or Brand Level?
There is no patent that clearly answers this question, but Google does indicate in its Search Quality Guidelines that E-A-T applies to “the creator of the main content; the main content itself, and the website.”
Google also edited the quality guidelines in 2019 to expand the notion of YMYL from page-level considerations to also pertaining to “topics.”
Google updated its guidelines on YMYL to expand from “pages” to “pages” and “topics”This phrasing throughout Google’s documentation suggests that Google’s E-A-T evaluations are likely done on the entity level.
This would make sense, given what Google is building with its Knowledge Graph: a repository of 500 billion facts about 5 billion entities, and the connections between them.
While Google’s patents haven’t explicitly mentioned evaluating entities based on E-A-T considerations, Gary Illyes from Google stated at the 2019 Pubcon conference that:
“We have entities for very popular authors, like if you were an executive for the Washington Post, then you probably have an entity. It’s not about the author, it’s about the entity.”
Therefore, just like how “YMYL” can be a topic within a page, a recognized entity within your page can potentially have good or bad E-A-T.
This is why taking steps like adding author biographies – especially for authors recognized in Google’s Knowledge Graph – can be an effective SEO strategy.
Beyond that, SEO pros should look for opportunities to ensure Google is able to easily identify entities on the page, such as using structured data in order to improve E-A-T.
For further reading, Google’s Related Entities patent from 2013 provides more context about how Google identifies and ranks entities.
Takeaways & Speculation
E-A-T is a bit of a gray area in SEO.
But gaining an understanding of Google’s patents over the years helps to solidify what appears to be a consistent mission by Google:
- To identify authors, organizations, and many other types of entities.
- To classify those entities and understand the level of authoritativeness required to be experts on those topics.
- To make connections between entities.
- To gain an understanding of how authoritative entities are on their respective topics.
- To determine the characteristics of highly authoritative entities and experts, and potentially use that knowledge in their ranking considerations.
Therefore, any steps you can take in your SEO strategy to streamline and facilitate this process with Google will likely result in stronger SEO performance.
E-A-T-focused SEO does not include quick fixes or short-term hacks.
It involves ensuring that all the best characteristics about your brand, your authors, and your experts are clearly displayed on your site.
Not only for SEO benefits but – more importantly – to instill a sense of trust in your users.
More Resources:
- E-A-T & YMYL: How to Boost Your On-Page SEO
- How to Improve Your Website’s E-A-T
- How Search Engines Work
Image Credits
Screenshot taken by author, October 2020
Sorry, the comment form is closed at this time.