See filter

Rank brain: Google AI and search rankings

Thu, 29/10/2015 - 13:44

Recent developments in Silicon Valley have seen the Internet powerhouse Google rebrand itself as a subsidiary of Alphabet, a newly formed Multinational Tech Conglomerate, and openly announce a wealth of subsidiary companies looking to change the world. Literally.

Just a few days ago, Bloomberg reported on the use of Rank Brain – Google’s AI search development – affecting 15% of searches using artificial intelligence to interpret search queries to serve the best results for users. Most algorithm changes affect around 1 to 4% of all searches internationally on first iterations, so 15% of international searches being affected by AI is sure to have an impact on what the face of optimisation looks like.

Machine learning has been at the forefront of Alphabet’s hidden strategy, with Google Brain being in development since 2011 as part of their exploration into deep-learning. Essentially, deep learning is training computers to recognise patterns and then interpret new unseen data and relate it to their existing consciousness, much like a human brain. While many of you have seen elements of this recently with Deep Dream from Google, where based on the patterns and elements in uploaded images, Google would relate often horrifying images to the user.

While historically search results from algorithms have been calculated based on previous search data, such as those seen in Google Predict (the queries that appear as you type in the search bar based on popular results), Rank Brain is set to interpret queries that are unusual and different from these, such as colloquialisms and interpret what would be best to serve the user. These operate as machine-generated guesses and through refinement of the technologies Google will be able to handle almost any query with a mythical accuracy. Currently, in the early phases, in comparison to search engineers eyeballing and guessing top ranking pages they were correct around 70% of the time. Rank Brain has already started to top the 80% accuracy in results.

Currently Rank Brain is considered to be the 3rd strongest ranking factor in regards to the results served to users every second around the world. What this means to SEO and marketing companies is that reference and citation is going to be become infinitely more important as you attempt to reverse engineer the way deep learning will affect search results. The more mentions and relations around a variety of terms, means that your site will be understood to relate to more searches and have stronger references. This will also mean that misspellings and colloquialism optimisation will become less viable as strategies over time with the interpretive nature of the AI recognising that users are looking for elements beyond the incorrect words. Gone are the days of optimising for three keyword phrases or building comment links, as deep learning pushes a further step into interpretive search where you will need your site referenced through quality content, on quality sites for Rank Brain to remember your site as being the best content for users.

These however, much like the initial responses by Rank Brain, are just our speculations in regards to the learned knowledge we have from search and data vectors on the internet. In time we’ll learn more and have a better answer for you.