Google Rankbrain was officially released mid-2016. This full algorithm release took most people in the SEO world by surprise, including myself. But, when we take a closer look, it seems that this machine learning (ML) algorithm was released in 2013 with the Hummingbird update and has been working behind the scenes ever since, collecting vast amounts of data to learn from.

As we have seen with Microsoft’s Twitter Bot Tay, ML can go horribly wrong. However, Google have been patient as to when to release this fully into the search landscape.

Rankbrain hasn’t received the coverage it deserves within the industry, and certainly hasn’t been seen for what it could be; the biggest shift in search since the move from directories to complex algorithms. The reason that this is so important is that Google’s understanding of semantic content and queries is now at a fairly accurate level. This understanding will constantly be evolving as the algorithm learns from new data, and therefore constantly improving.

The examples below show a Google search that has displayed a high level of understanding, not only of the keywords in my query, but also the intent behind search.

In the example the “funkyityfootoo” auto complete suggestion seems out of place, and it is from a keyword perspective. But, if you click on this auto complete for this weird listing, the search engine result page (SERP) returns results that are in fact based on the original query – specifically an experiment run by an SEO on the indexing of accordion content using appropriately niche terms. No matter how irrelevant the query appears, it seems that it can still be (correctly) semantically related!

RankBrain Semantic Understanding

Google has been returning auto completes previously for easier queries, but the level of understanding displayed in the above query indicates that Google is ramping up it ML efforts.

There was also talk within SEO circles of Google using the disavow function in GSC to collect large amount of data from webmaster on domains that are considered spam, and using this to seed a ML element of Penguin. This is something that is not out of the realms of possibility.

Soon we will see more ML elements baked into the core algorithm. For example, ML can be used in backlink metric calculations, applied to weighting for anchor text on keywords and even a thematical understanding to domain and assign value based off domain thematical relevance (which is something being done now to a small extent).

I think of ML as a snowball that moves fast and grows even faster. As SEOs we need to be cognisant to the paradigm shift that will be coming and how we deal with new challenges as a result. If you can take one key learning away from this it would have to be that ML is not only being used by Google, but ML is being used by Google successfully. The death of the pure keyword-focused approach has been cited for a while now, in favour of topical relevance and natural language use, but here we can see some of the clearest evidence yet that Google really is moving beyond keywords and into real understanding of normal language. Plan your strategy accordingly!

This is just one of the reasons why 4Ps isn’t just another SEO agency. To discuss how your brand could use user-driven query targeting and technical SEO knowhow to boost your bottom line, give us a call on +44 (0)207 607 5650 for a no-obligation coffee and chat about data, marketing and technology across all inbound channels. What could a 4Ps search marketing consultant do for you?