Google’s Hummingbird - no big deal, for now
Google has announced what has been a silent shift to a new search algorithm, Hummingbird, the the biggest overhaul to its search engine since "Caffeine" in 2009, affecting 90% of searches.
Google launched the Knowledge Graph, and the associated knowledge cards, last year. They are the special results that appear when searching for things like "things to do in paris", and demonstrated that Google understands, to some extent, the meaning of a search term and therefore is better able to return results for it.
The unveiling of Hummingbird (incidentally the first time I know of that Google has given a name to its entire algorithm), shows Google has taken the algorithm behind the Knowledge Graph and applied it to the entire index instead of just the Knowledge Graph.
It's the next step in Google's quest for artificial intelligence.
Practically, this seems to mean the Google algorithm now considers more carefully full sentences rather than having emphasis on individual words.
That's quite a big departure from how Google is believed to have worked in the past where seemingly inconsequential words were not included in the "forward index" that search engines build for each page they crawl to enable them to retrieve and sort documents faster. In a crude example, a forward index may drop the word "the", meaning the two searches "the warmth" and "warmth" would end up being treated very similarly. A more advanced algorithm and index would actually recognise the importance of "the" in this instance since it determines whether the user is searching for a song or the concept of warmth and, most likely, heating.
Of course Google pre-Hummingbird was already advanced enough to differentiate between these two searches, but it does serve as a sort of conceptual example that, if you scale up in complexity, gets close to the difference between the old algorithm and the new one.
I wouldn't get too excited about this being a massive change that is going to start causing huge fluctuations in rankings. Google said it started using it about a month ago (based on observed rank fluctuations from Greenlight's Integrated Search Marketing technology platform, Hydra, our best guess is August 20th) so unless it has started small and is planning on ramping it up later, which is somewhat possible, the impact has already happened.
And, while the algorithm has now been given a name, it's still largely the same algorithm under the hood, albeit with some tweaks and additional components. So all the traditional stuff (e.g. PageRank) and not so traditional stuff (Panda, Penguin), is still in the mix and quite as relevant as it ever was. For the foreseeable future I think the major fluctuations that keep marketers up at night are still going to come in particular from Panda and Penguin updates, rather than the changes launched in Hummingbird.
But the way Hummingbird is said to work does have some obvious implications for SEO. Looking at full sentences, instead of just words on a page, means some trusty approaches to optimisation, particularly optimising "automatically" by having repetitions of keywords on pages by virtue of their appearance in headings, titles, navigation elements, footers etc… aren't going to be as effective. Keywords in those elements often lack context, and instead Google might want to see content with meaning, which comes from the other words nearby and full sentence structures. It might be impossible to escape the implication that pages just have to have proper written content.