Perhaps the most striking event of the last months in the field of SEO was to discuss BERT algorithm, which began after members of Google ads that the algorithm is already being used in the search for a number of requests. Although, in my opinion, this algorithm does not have much of any appreciable effect on the traditional areas of interest in the search seoshnyh, ignore it, I think it would not be entirely correct.
In the developer blog Google search engine appeared October 25, 2019, the year the article under the title «Understanding searches better than ever before» ( « The understanding of search queries is better than ever before") authored by Vice President of Google Search Pandu Nayak (Pandu Nayak ). In addition it is very lengthy racskaza on search technology called BERT, first announced a year ago, has been mentioned about it has been launched in the "battle" Search: «We're also applying BERT to make Search better for people across the world».
The article as a fact are considered examples of improvements in the issuance of some queries. However, neither of which separate release does not explicitly been stated, and it looks that the improvements introduced gradually for some time, and SEO-community had not noticed them before the release of this article. But after the statement appeared on the release date of the algorithm. Thus, the first known Western SEO-quarters Barry Schwartz Specialist (Barry Schwartz) in hot pursuit in the day of the article release stated that the algorithm was the roll out this week, and a few days later another expert George Nguyen (George Nguyen) called the exact date : October 21, which is then rastirazhirovali various industry media.
However, do not notice a change in the search really is not surprising, even given the fact that Google experts estimate that they will affect every tenth request in English. The fact that the improvements relate to the long tail mikrochastotnyh queries in natural language, which simply drops out of sight seoshnyh issue of monitoring systems. In connection with this recall yandeksovskie releases algorithms " Palekh " and " Queens ", which are also almost did not affect the traditional areas of SEO-industry interests in the SERP, though Yandex employees, of course, gave them great value. Apparently, the same thing happens with the consequences of the introduction of BERT algorithm.
So SEO-experts, in my opinion, can relax and to follow the advice of Evangelist Google search Dann Sullivan (Danny Sullivan), given to them in his Twitter account: «There's nothing to optimize for with BERT, nor anything for anyone to be rethinking. The fundamentals of us seeking to reward great content remain unchanged. "(Loosely translated -" There is nothing to be optimized for BERT, as well as something to rethink the fundamentals of our search for a decent content remain unchanged "..). Worrying, perhaps, is only a "cautery" semantics for infozaprosam who should think about improving the quality of their articles "on all occasions."
However, any changes in the search for many site owners will now associate it with the algorithm BERT. Revealing a curious dialogue , which took place during the meeting time video for webmasters Google employee John Mueller (John Mueller). The owner of the site, which has fallen on November high-demand on page 4-5, while maintaining the position of the low-frequency tail, blamed the fall BERT algorithm. Thus showing a complete misunderstanding of the nature of the algorithm, and having received from Muller's expected answer, that the situation of his website linked to routine changes in the algorithm, non-BERT technology.
I believe that if a position in Google SERP and any other site significant changes have taken place in November, it is very unlikely that this responsible BERT.
As for the Russian-language queries, it is still not clear how broadly applicable to them BERT algorithm and whether it is used, although Pandu Nayak in his article mentions that he used to improve search around the world.
As one example of the algorithm Pandu Nayak leading English-language query «parking on a hill with no curb », demonstrating how to improve the delivery of considering the word "no", before to ignore:
Let's see what happens in the issue of Russian-speaking counterparts of this request. Query "car park on the side of the curb if not" handled properly, the first two snippets we see that the documents must contain the answer to the question:
However, just by changing the query to "parking on a slope without border", we see that the snippets of the first places of issue no longer point to the fact that the document contains the answer to the question:
By the way, Yandex and his "Queen" solves this problem similarly. In the first case snippets second and third places contain the answer to the question:
But in the second case - no:
In general, the algorithms for processing queries in natural language in both search engines clearly still have work to do.