Skip to main content

Google Search: From PageRank to RankBrain

Until recently, Google’s PageRank algorithm was determined by hundreds of factors chosen by engineers at Google without the use of machine learning. The basic premise of PageRank is that it allows sites to “vote” with links. More specifically, the most popular sites have votes with higher weights, while less popular sites have weaker votes. This voting system allows for the most popular results to appear for a given set of keywords. However, this has not always been a fool-proof method.


To combat ambiguous queries that can lead to “Search Engine Fatigue,” the inability to find what you are looking for, Google has implemented RankBrain. Previously, when a search included colloquial terminology, unfamiliar terminology, or was phrased naturally (i.e. as one person would talk to another), PageRank performed suboptimally. When Google Search does not recognize the terminology you are using, RankBrain now allows for Google to “guess” what you mean.


So how does RankBrain guess what you mean? By using word vectors, RankBrain can compare words that are linguistically similar. In the process, words that would have previously delivered inaccurate results are mapped to vectors that allows Google Search to infer what you mean. More natural searches like “what is the best coffee shop in Ithaca” are now mapped to similar, more popular searches like “best coffee shop Ithaca.” To further increase the quality of results, Google records search terms and results online and offline to train the RankBrain algorithm. Ultimately, RankBrain has become one of the most important factors in determining search results, though the fundamentals of PageRank are the same as they were before the new methodology was introduced.


My discussion of RankBrain above is in response to the following sources”



My mention of “Search Engine Fatigue” originates from the following source:



Leave a Reply

Blogging Calendar

October 2017
« Sep   Nov »