Google Rank Brain

RankBrain – Google’s Mysterious AI Resource

Google has come a long way from where it started. Users have shifted from desktop to mostly mobile queries. Google has gone from being a stripped-down search platform to being an indispensable platform for business, and it has been a technology innovator the whole way, but some of it’s most important elements are practically invisible to users and advertisers. Google has always deployed major changes to its internal code with large algorithm updates, generally named after animals. Google’s updates often deal with closing loopholes and improving the user experience. 

Birth of RankBrain

First created in 2015 RankBrain isn’t an AI in the typical science fiction meaning, but is more accurately a deep machine learning algorithm which utilizes word vectors to sort and weigh data on the fly, often in ways that were not originally conceived by its creators. 

The machine learning algorithm is constantly updating its pool of words and how they combine to create different intent. The algorithm can then weigh a response to a query against the likelihood of producing the desired result, which in the case of Google is generally giving the user what they were looking for, even if they didn’t give enough information to determine that. 

Some searches can be parsed based on data outside the search, for example, if you search for Italian food on your phone Google can usually see where you are and will assume you are looking for a place to eat, and will serve up those results first, based on the device type, location, time of day, etc.

But what of connections and interactions that aren’t as obvious? With 5.6 billion queries a day (Source SEOTribunal.com) Google has an abundance of activity to monitor and base predictions on. By leveraging that data with a machine learning algorithm Google improves the relevance of its results, and therefore it’s bottom line. 

Where Does RankBrain Fit into the Search Results

When it was first launched RankBrain was used as a “signal” in 15% of searches. Signals are factors in determining what will be in the Google search results (SERP) what their position will be and how the results will be presented. Once RankBrain’s effectiveness was seen this was expanded. Generally, a straightforward and common request will not require RankBrain, but when the language is less common, or words are combined in a phrase that doesn’t trigger a common search result RankBrain is used to help make an educated guess. 

Neural Matching

While RankBrain is the most ambitious part of Google’s current algorithm in terms of machine learning there are other elements that help improve relevance when the desired search result may be unclear. Neural matching creates links between phrases and their potential intent. For example “Why does my TV look strange” is parsed as “Soap Opera Effect” based on the analysis of previous user behavior (Source searchenginejournal.com).

What goes into the Search Result

Google doesn’t use one source or list to determine what you are seeing in the search results. Each of these elements (user history, RankBrain, Neural Matching, etc) are signals that Google weighs and combines to determine the position and layout of the search results to best match a user’s intent. 

Implications for SEO

What does this mean for businesses wanting to show up on the search results and drive qualified users to their site? In short, the answer is to design your site based on the user experience. Use language your customers will use and write for humans, rather than for search engines. Strong relevant content will help show that you are a valid result for the query and reassure users when they click that link in the results. 

Structured Data is also important for ranking well. Does your site follow schema markup guidelines? Make sure addresses, products, phone numbers, etc. are called out in a way that they can be read and accurately interpreted by Googles indexing algorithm.

What RankBrain Doesn’t Look At

RankBrain doesn’t monitor and factor in all elements of user behavior, for example, it will not monitor how long someone stays on a page, bounce rate, or CTR. For purposes of SEO, these are also called on-page signals, and while a huge part of what determines SERP, they aren’t a factor for RankBrain.

Position Versus Inclusion

RankBrain is part of the overall results, but it is not a positional factor. Rather it allows a result that might not be indicated based solely on keywords to be included in the search results. The actual position and how the result are displayed is defined by other parts of the algorithm.

A Final Thought on SEO for the Google Search Algorithm.

Algorithms will continue to change, but a good rule that will stand the test of time is to make your content for people and avoid anything that google could interpret as an attempt to fool or mislead the algorithm. Google can delist a site that they feel is employing “Black Hat” SEO tactics. They are under no legal obligation to allow it back on the search engine results.