Google Search Ranking Core System Underlying Technology Guide

Google uses automated ranking systems that look at many factors and signals about the hundreds of billions of web pages and other content in Google’s search index to deliver the most relevant and useful results, all within seconds.

Google regularly improves these systems through rigorous testing and evaluation, and sends out notifications when the ranking system is updated (and if those updates might be helpful to content creators, etc.).

This article looks at some of Google’s more well-known ranking systems. Some of Google’s core ranking systems are described, which are the underlying technologies that generate search results for queries. In addition, some of the systems that address specific ranking needs are described.

You can also visit Google’s How Google Search Works site to learn how Google’s ranking system works in concert with other processes to enable Google Search to fulfil Google’s mission of integrating the world’s information for the public’s use, for the benefit of all.

BERT
Bidirectional Encoder Representations from Transformers (BERT) is an AI system used by Google that allows Google to understand how different combinations of words convey different meanings and intentions.

Disaster Information Systems
Google has developed several systems to provide timely and useful information in the event of a disaster, whether it is a personal crisis situation, a natural disaster, or other widely spread disaster situation:

Personal Crisis: Google’s systems work to understand when people are seeking information about a personal crisis and display the contents of hotlines and trusted organisations in the case of specific queries about suicide, sexual assault, poisoning, gender violence or drug addiction.

SOS Alerts: During natural disasters or widespread calamities, Google’s SOS Alerts system attempts to display information issued by local, national or international organisations. This information may include emergency phone numbers and websites, maps, translations of useful phrases, donation opportunities, and more.

Duplicate Information Removal System
When searching on Google, you may see thousands or even millions of matching pages. Some of these parameters may be very similar to each other. In such cases, Google’s system displays only the most relevant results to avoid unhelpful duplicate information.

Google also considers the Featured Summary when removing duplicates. Even if a page detail is promoted to a Featured Summary, Google will not repeat this listing on the first page of search results. This helps to organise search results and helps users find relevant information more easily.

Exact Match Domain System
Google’s ranking system looks at the words in a domain name as one of many factors in determining whether content is relevant to a search. However, Google’s exact match domain system ensures that it doesn’t give much consideration to content hosted under domains created specifically to be an exact match for certain queries. For example, a user may create a domain name containing the words “best restaurant for lunch” in the hope that all of those words in the domain name will improve the ranking of the content, and Google’s system will adjust accordingly.

Updating the system
Google has various “query should be kept fresh” systems designed to display expected fresher content for a query. For example, if someone searches for a film that has just been released, they may want the latest reviews, rather than the oldest reports since the film’s production began. As another example, a typical search for “earthquake” might return content about earthquake preparedness and resources. However, if there has been a recent earthquake, then news stories and newer content may appear.

Useful Content System
Google’s Practical Content System is designed to ensure that users see original, practical, user-written, user-facing content in search results, rather than content that is primarily intended to drive search engine traffic.

Link Analysis Systems and PageRank
Google has a variety of systems to understand how web pages are linked to each other to determine what is relevant and which pages are likely to be most relevant to a query. This includes PageRank, which is the core ranking system used by Google when it was first released. For those interested, the original PageRank research papers and patents are available for details. Since then, the way PageRank works has changed significantly and remains part of Google’s core ranking system.

MUM
The Multitask Unified Model (MUM) is an AI system capable of understanding and generating language. It is not currently used for general ranking of Google searches, but for some specific applications, such as for improved searching of COVID-19 vaccine information and improved labelling of selected abstracts.

Neural Matching
Neural Matching is an AI system that Google uses to understand the representation of concepts in queries and web pages and match them to each other.

Original Content System
Google has systems in place to ensure that original content (including originality reports) is displayed prominently in search results and that they are ranked ahead of cited content. This includes support for special canonical markup that creators can use to help Google better understand which is the primary page if there are duplicate versions of the page in multiple locations.

Removal-Based Demotion System
Google’s policy allows for the removal of certain types of content. If Google processes a large number of such removal requests involving a specific website, Google will use this as a measurement to improve Google’s search results. Specifically:

Legal Removal: If a large number of valid copyrighted content removal requests involving a specific website are received, Google will accordingly lower the ranking of other content on that website in its search results. This way, if other infringing content exists, users are more likely to see the original content rather than the corresponding infringing content. For complaints involving defamation, counterfeit products, and court-ordered removals, Google uses a similar measure of downgrading.

Personal Information Removal: If Google handles a large number of personal information removal requests involving a site that employs paid removal practices, Google will lower the rankings of other content on that site in its search results, and will seek to learn whether other sites engage in similar behaviour; if so, it will take a downgrade measure for content on such sites. Google may take similar demotion practices for websites that receive a large number of human-searchable content removal requests. In addition, Google has put in place automatic protection measures to prevent explicit personal images posted without the consent of the person concerned from being ranked higher in queries involving the name in question.

Web Experience System
Users prefer websites that provide a good web experience. Therefore, Google has developed the Web Experience System to evaluate various criteria, such as the speed of loading, the suitability for mobile devices, the absence of intrusive inserted advertisements on a web page, and the safety of the web page presentation. If there are multiple possible matching pages with relatively consistent relevance, the system will prioritise content with a better web experience.

Paragraph Ranking System
Paragraph Ranking is an AI system that identifies parts or “paragraphs” of a web page to better understand how relevant the page is to the content being searched.

Product Rating System
The product rating system is designed to better reward quality product reviews that contain insightful analyses and original research, and are written by experts or enthusiasts who are knowledgeable about the subject matter.

RankBrain
RankBrain is an AI system that helps Google understand the relationship between words and concepts. This means that even if content doesn’t contain all the exact matches of words used in a given search query, Google can understand how relevant that content is to other words and concepts, so it can better return relevant content.

Reliable Information Systems
Multiple systems display the most reliable information possible in a variety of ways, for example to help present more authoritative pages and lower rankings for poor quality content, as well as to boost rankings for quality news. If there may be a lack of reliable information, or if Google’s systems are less confident about the overall quality of the search results, Google’s systems will automatically display content suggestions about rapidly changing topics. These content suggestions will give you tips on how to find search results that may be more useful.

Website Diversity System
Google’s Diversity of Sites system ensures that Google does not generally display more than two web page product details from the same website in the top search results, so that no single website dominates the top search results. However, if Google’s system discovers that more than two product details from the same website are highly relevant to a particular search, it may still show more than two such product details. Web diversification systems often treat subdomains as part of the root domain. For example, the system will treat the subdomain (subdomain.example.com) and the root domain (example.com) as part of the root domain.