All algorithms work for all languages

How the search algorithms work

With the amount of information on the web, it would be almost impossible to find what you are looking for without pre-sorting. That's why Google's ranking systems exist: They search billions of websites in the search index and present you with the most relevant and useful results in a matter of seconds.

These ranking systems are based on a number of algorithms. A variety of factors are used to get the most useful information. This includes the words used in your search query, the relevance and usefulness of pages, the expertise of sources, and your location and settings. The weighting of the individual factors depends on the type of your search query. For example, the topicality of the content plays a greater role in answering questions on current topics than in dictionary definitions.

To ensure that the search algorithms meet the high requirements for relevance and quality, we have a strict procedure that includes both live tests and thousands of trained external evaluators to assess the quality of Google searches. These evaluators follow strict guidelines that set out our goals for search algorithms. These guidelines are publicly available and can be viewed by anyone.

Below you will find more information about the key factors that we use to decide which results you will get for your search query:

  • Word analysis

    Word analysis

    In order to provide you with useful answers, we must first understand exactly what you are looking for. Therefore, in a first step, we analyze the meaning of the words in your search query. We create language models to decipher which words to look for in the search index.

    These language models, incorporating the latest research into understanding natural language, can perform simple tasks such as spelling mistakes. But they can also solve more complex problems and, for example, identify the type of search query. With the help of our synonym system, Google search can, for example, recognize exactly what you are looking for, even if the word entered has multiple meanings. This system took more than five years to develop and it improves the results of more than 30% of searches.

    We also try to identify the category of information we are looking for. Is it a very specific term or is it more of a general question? Does the search query contain words such as "review", "pictures" or "opening times" that provide further information? Is the user looking for trending topics and would like to receive the latest results from that day? Or is he looking for a nearby company and needs information about his area?

    A particularly important dimension of this search query categorization is the analysis of whether you are looking for new content with your query. When you search for trending words, our timeliness algorithm interprets this as a signal that current information may be more useful than older pages. This means that you get the most up-to-date information when you search for the latest "Bundesliga results", "Let's Dance" or "VW earnings".

  • Matching the search term

    Matching the search term

    In the next step we look for websites that match your question. Our algorithms first search the search index for the term to find the right pages. They analyze how often and where the keywords appear on a page, for example whether they are in the title, in headings or in the main text.

    The easiest way to tell whether information is relevant is when a website contains the same keywords as your search query. If these keywords appear on the page, in the headings, or in the body of the text, it is even more likely that the information is relevant. In addition to simply checking keyword matching, we use aggregated and anonymized interaction data to assess whether search results are relevant to specific queries. We convert this data into signals that our machine-learning systems can use to better assess the relevance.

    In addition to matching the keywords, the algorithms look for clues to determine how relevant the results are to the search query. If you search for "dogs", you are probably not looking for a page that has the word "dogs" appearing multiple times. So we try to find out whether a page contains an answer to your question or just repeats the search term. The search algorithms analyze whether pages contain relevant content - such as pictures or videos of dogs or a list of different dog breeds. Finally, we check that the page was written in the same language as your search query. The results in your preferred language will then be displayed near the top of the search results list.

    It is important to note that while our systems look for quantifiable signals of this type to assess relevance, they are not designed to analyze subjective concepts such as the point of view or the political orientation of the content of a page.

  • Ranking of useful pages

    Ranking of useful pages

    For a common search query, there are thousands or even millions of web pages with potentially relevant information. In order to display the best pages first, our algorithms first have to determine how useful each web page is.

    These algorithms analyze hundreds of different factors in order to find the best information on the web - from the topicality of the content to the frequency of the search term on the page to the user-friendliness of the respective website. In order to determine the trustworthiness and competence for a certain topic, we look for websites that other users prefer for similar questions. If other well-known websites link to a page on the same topic, that is a good sign that the information fits well there.

    Unfortunately, there are many spam websites out there trying to sneak their way to the top of search results using techniques like repeating keywords or buying links that give PageRanks. These websites are user-unfriendly and can mislead or even harm Google users. Therefore, we write algorithms that capture spam and remove websites that violate Google's Webmaster Guidelines from our search results.

    The content on the Internet and the general information system are constantly changing. We therefore continuously measure and evaluate the quality of our systems in order to strike the right balance between relevance and reliability of the information and thus ensure our users' trust in the results.

    additional Information

  • The best results

    The best results

    Before we present you with search results, we analyze whether the information fits together: Do the search results relate to one topic or to several topics? Are there too many results for a narrow interpretation of the search term? We try to present as much different information as possible in the most appropriate form. As the web continues to evolve, our ranking systems also evolve to provide even better results for even more searches.

    These algorithms analyze signals that indicate whether all of our users can see the result. For example, it is checked whether the website is displayed correctly in different browsers; whether it is designed for all device types and sizes, including computers, tablets and smartphones; and whether the page load times also work for users with slow internet connections.

    Because website owners can improve the usability of their website, we endeavor to notify them in advance of any significant, actionable changes to our search algorithms. For example, in January 2018, six months before the changes came into effect, we announced that our algorithms would soon take website page speed into account. To help website owners, we've provided detailed guides and tools, such as PageSpeed ​​Insights and Webpagetest.org, so they can see what they might need to adjust to make their websites mobile.

    You can find more information about the tools and tips that Google provides website owners with here.

  • Contextual reference

    Consideration of the context

    We can use information like your location, previous search history, and search preferences to provide you with more relevant and useful results.

    So that we can show you relevant information, we take your country and location into account. For example, if you search for "football" in Germany or the United States, Google will likely come up with results for American football first. But if you are in Great Britain and search for "football", the ranking for football is higher. The search settings are also an important indicator of the right search results, for example the preferred language or the activation of SafeSearch, a tool that filters out objectionable search results.

    In some cases, we also personalize the results by taking into account your recent searches. For example, if you search for "Barcelona" and recently entered "Barcelona versus Arsenal", you are probably looking for results on the football club rather than the city. At myaccount.google.com you can specify which search queries can be used to improve the results.

    Google Search also includes some features that personalize your search results based on activity in your Google Account. For example, if you search for "events near me," Google can come up with some recommendations for event categories that we think you might be interested in. These systems are used to deliver results that match your interests. They are not intended to be used to infer your skin color, your religion or your political views.

    At myaccount.google.com you can specify which search queries can be used to improve the results and which data is saved in your Google account. Deactivate "Web & App Activities" to deactivate the personalization of Google Search based on the activities in your Google Account.

Organization of information

Back

Useful answers

Further