Yandex filters & algorithms. 1997-2018
In this article you will find all the changes of Yandex algorithms, starting from the day of its birth to the present day. This article will be updated.
In the beginning I’d like to highlight only the most important, in my opinion, changes of algorithms that you should know about if you plan any SEO activities on Yandex.
Key facts about Yandex algorithms, filters and penalties
Introduction of local SEO. Starting this year the website promotion in all regions of the country is becoming more and more difficult.
Introduction of the AGS algorithm, aimed at reducing the visibility of websites without added value (including sites created only for selling links). AGS penalty then had several versions.
MatrixNet is the first Yandex algorithm based on machine learning (details).
Introduction of behavioral factors (CTR in search results, bounce rate, activity on the website etc.). Along with that, SEO market introduces services for manipulation of behavioral factors.
Spectrum algorithm — in case of unclear or ambiguous queries (like pizza or apple) the search results contain sites of different categories (forums, directories, online shopping, etc.), to fully answer the query (details).
First sanctions for the manipulation of behavioral factors.
SERP personalization — each user sees the search results, based on his/her interests and requests (details).
Introduction of usability (ease of navigation, no disturbing ads) as a ranking factor.
A separate group of factors for commercial sites: prices, number of different items, terms of delivery, etc. (details)
Links have been disabled as a ranking factor.
Return of links as a ranking factor, but a significant change in their influence on rankings.
Introduction of Minusinsk — penalty for unnatural (purchased) links .
All changes in algorithms of Yandex search engine in 1997-2016
Sep 23, 1997 – the official birthday of the company “Yandex”. Search engine could already take into account the morphology, the distance between words and relevancy of the document (webpage) to the specified query.
2005 — Nepot filter (‘nepotism’)
The first Yandex sanctions as a response to the growing of link exchange, which algorithm could not cope with. On search results page (SERP) you could see such nonsense as, for example, Microsoft website for the query “hemorrhoids”.
Nepot filter punished websites that published links to quickly get into the Top-1. If sanctions are imposed on the links, they completely lose their link juice.
2006 — Automatic geo classification of websites.
The user is able to specify that he is interested only in websites of his region.
Jul 2, 2007 — Change in ranking formula
Different ranking formulas for single and multiple words queries.
For the first time Yandex announced its algorithm adding a post on the SearchEngines.ru forum. In addition, since that time webmasters have the opportunity to contact Yandex search team support. First link networks appeared. SEO specialist’s work is mainly link buying.
Aug 7, 2007 — New ranking formula for single word queries.
The main pages get priority for popular single word queries.
Jan 17, 2008 — “8 SP1″.
The first Yandex algorithm, which got its own name. Since that time Top-10 of search results is mainly occupied by old and trusted websites. The term TrustRank has been introduced, that is, the credibility of the website. The power of links from the main pages is decreased. SEOs massively change their strategy to buy links from internal pages.
Mar 18, 2008 — the continuation of the offense on SEO links.
Websites, that occupied top positions using links from the main pages, lose their positions. Links from internal pages still have SEO impact.
Apr 4, 2008 — Foreign documents apear in the search results.
Yandex starts to search on foreign internet, dumps English websites appear in search results. At the moment the size of foreign internet database is about 1 billion websites
May 16, 2008 — Magadan
Yandex has learned to decipher abbreviations and transliteration. The number of ranking factors increased twice.
This was the first algorithm which had the test version of search results available on buki.yandex.ru, where within a month webmasters could check out the results and leave their feedback for Yandex team.
July 2, 2008 — Magadan 2.0
Added new factors that took into account the uniqueness of content, and began to classify queries into commercial/non-commercial and location-dependent and location-independent. The value of website age has been reduced.
Sep 11, 2008 — Nakhodka
Increased value of the internal pages in the search results. For medium popular and some high popular queries internal pages appeared in SERP along with the main pages.
The subsequent morphological changes, search results differ in case of query in the singular and in the plural.
Informational pages, in particular, Wikipedia, appear in the search results for commercial queries. Fight with cloacking is increasing.
Apr 10, 2009 — Arzamas. Local SERP (details)
SERP now depends on geography. The option “show websites only from my region”, that a user should enable manually back in 2006, Yandex now enables automatically. Now search engine considers the user’s locale by IP address. The term regional (local) SEO appears. Regional SERP introduced in Moscow, Saint-Petersburg, Ukraine, Belarus and Kazakhstan. Setting website region appeared in Yandex.Webmaster panel.
Apr 30, 2009 — Fight with aggressive advertising begins
Yandex started to lower ranking of sites using popunder or clickunder ads.
Jun 24, 2009 — Arzamas 1.1.
Improved ranking formula for the regions of Russia, except Moscow, St. Petersburg and Yekaterinburg.
Aug 20, 2009 — Arzamas 1.2. Classifier of query location-dependence
Old and trusted websites get priority in search results for location-independent queries. Mostly main pages appear at the Top of SERP for location-independent queries.
Sep 28, 2009 — AGS 17
The first official mention of AGS filter, which, according to Yandex, has been working since 2006. After getting under this filter, only 1-10 pages of punished domain remain visible in search results. The first version of this filter mainly touched automatically filled pages and websites with non-unique content.
Nov 17, 2009 — Snezhinsk. The first algorithm based on machine learning method named MatrixNet.
Informational websites, but not commercial, become leaders in the Top results for highly popular single word queries. Quality of the whole domain begins to play a greater role than relevancy of each page.
Official birth of MatrixNet, a new method of machine learning. The number of factors that are considered in the evaluation of documents rose to several thousand. Behavioral factors appeared in the ranking formula.
Yandex has learned to distinguish commercial and non-commercial (natural) links using the anchor analysis (see the report).
Dec 18, 2009 — AGS 30
Increased the number of factors deciding whether to exclude a website from the index. Filter punishes sites with low quality content, which, on the one hand is unique and on the other, has no added value for users.
Filter can be applied both manually and automatically.
Dec 22, 2009 — Konakovo
The local ranking now works in 1250 cities of Russia.
Jan 20, 2010 — Anti-“footcloth”
The name comes from the definition of very long and over optimized texts for selected keywords. Filter punished websites for such a over optimized texts. Filter affects the individual web pages and treated by reducing the text length several times (but keyword density could remain at the same level). Filter anti-“footcloth” is a precursor to the algorithm “You are spammy”, which was released in 2012.
Mar 17, 2010 — Snezhinsk 1.1
The update of ranking formula for location-independent queries. Priority in search results is given to non-commercial websites — encyclopedias, reviews.
Sep 13, 2010 — Obninsk
Improved detection of web sites region. In the search results for location-independent queries, the advantage is given to web sites, that have no region set in Yandex.Webmaster.
The algorithm had a negative impact on SEO techniques that used low quality links. Using the time factor for commercial links, which means that the publisher begins to transfer link juice to the acceptor only in 2-4 months after link was published.
Nov 2, 2010 — Improvement in ranking of author texts
Changes in algorithm for determining authorship of a document.
Dec 15, 2010 — Krasnodar. Introduction of “Spectrum”
Search results for popular queries become more diverse. In the Top 10 for a common query you could find online store, encyclopedia, forum, etc. (Spectrum technology). Queries are divided into 60 semantic categories (topics). Spectrum calculates proportions in which the answers on this topic should be presented in search results. The results are ranked so their spectrum is aligned with the spectrum of intents.
Thus, Yandex maximizes the likelihood that user will find exactly what he or she was looking for. (see the details of this technology)
Priority is given to behavioural factors (user actions on the search results page).
Yandex begins to index Vkontakte. Now it is possible to find user social profiles through the search engine.
Dec 17, 2010 — New ranking formula for location-dependent queries
In each location websites associated with the same region got privilege. Promoting one website in several locations becomes problematic.
May 2011 — Sanctions for behavioral factors manipulation
The first wave of manual penalties for websites of known Russian SEO agencies and their clients. On May 23 the official warning appeared on Yandex blog.
Aug 17, 2011 — Reykjavik
Search results now based on user’s language preferences. The first step towards SERP personalization at the level of a specific user.
Sep 13, 2011 — “You are spammy” filter
Penalties for spammy over optimized texts. Filter took behavioral factors into account: only those spammy texts that had low behavioral performance lost their rankings. Initially, filter was query-dependent and touched a single page.
Oct 5, 2011 — Yandex begins to consider website usability
Websites abusing ads automatically loose their rankings. Pages with awkward navigation begin to have problems with SEO.
Nov 23, 2011 — New algorithm for commercial websites (details)
Formula was supplemented by commercial factors: service usability, number of different product items, prices, shipping, additional services, etc. Initially, algorithm was run for the Moscow SERP, then expanded to other locations.
Dec 12, 2012 — Kaliningrad. Personalization of search results
Attempt of adaptation of search results and suggestions for each individual user, based on history of his queries and behavior on SERP. Promoting young sites in the personalized search results with conventional methods became difficult.
Yandex began its collaboration with Twitter. After Twitter opened up its database, the indexing speed of a website with regular tweets has improved greatly.
February 2013 — Yandex begins to mark in search results if website is infected with a virus or contains malware code.
Feb 7, 2012 — Personalized search suggestions for registered users
Yandex starts to take into account the history of user queries in suggestions in the search bar.
Feb 16, 2012 — Local search results for Yandex.Images
Different of image search results for Russia, Ukraine, Belarus and Kazakhstan.
May 15, 2012 — Penalties for sites with fake pop-up elements
Sites with pop-ups resembling system notifications or private messages from social networks, began to automatically fall in rankings.
May 16, 2013 — Announcement of the platform “Islands”
Yandex announced new platform “Island”. It was supposed to make a radical change in the format of the search results page where user could solve his problem (purchasing goods and services, hotel reservation, etc.) directly on the search page without going to the website.
However, for unknown reasons, the platform implementation was constantly postponed. On July 26, 2013 test version of “Islands” was introduced on Beta.yandex.ru.
November 6, 2013 — AGS 40
Features of low quality content have been updated. The main reason for filter punishment is the lack of added value on website. Filter mainly affects MFA sites (Made for Adsence), as well as sites for earnings on affiliate programs. AGS is automatic filter.
Dec 5, 2013 — Announcement of ranking without links
Head of Yandex search, Alexander Sadovsky, at the conference in Moscow reported that Yandex is preparing a new ranking formula that doesn’t take links into account. Execution of the algorithm promised in early 2014 and only for commercial queries in Moscow.
Mar 14, 2014 — Disabling links as a ranking factor
Formula update for commercial queries in Moscow in branches like real estate, tourism, electronics and household appliances.
Mar 20, 2014 — Reducing the visibility of pages with shocking ads
Sites with disturbing and shocking advertising began to fall in search rankings. Filter affected websites, which had ads with erotic pictures, clips from horror movies, photos of body parts with overweight, and other nasty ads.
Apr 15, 2014 — Changes in the AGS algorithm
Now in case of getting under AGS filter the number of pages in index does not decrease. Symptom of AGS is now the decline of TIC (Thematical Index of Citation) to zero.
Summer 2014 — Changing the formula for commercial queries
Expansion of list of branches where links are disabled as a ranking factor. Changing the queries classifier.
Sep 14, 2014 — Changing the formula of informational sites ranking
Tightening algorithm for non-commercial sites. Next wave of reducing visibility of web pages without added value.
Sep 19, 2014 — Penalties for aggressive advertising
Tougher sanctions against websites abusing ads. A large number of quality sites got sanctions mistakenly, where “regular” ad units were replaced with aggressive ads using viruses. Later this error was corrected.
Dec 1, 2014 — Sanctions for manipulation of behavioral factors
Sanctions affected thousands of websites, which manipulated clicks from the search results. For the first sanctions also affected sites that used so-called “links strengthening” service, which is offered by Russian links networks. Service offers to buy additional clicks on purchased links to “increase” link juice that is passed.
Dec 30, 2014 — Next wave of sanctions for manipulation of behavioral factors
In addition to websites that really used these services for manipulation, sanctions also affected websites from the so-called “human shield”, that is a list of websites on which the holders of these services have manipulated behavioral factors (CTR in search results, bounce rates, pageviews etc.). “Human shield” was used in order to confuse the Yandex algorithm. Site owners, of course, didn’t know about this fact.
April 15, 2015 — Announcement of algorithm Minusinsk
At the conference “ByNet Week” in Minsk Alexander Sadovsky, Head of Yandex Search, announced the introduction of a new algorithmic filter — Minusinsk, which purpose is to fight against link spam. After the announcement 9 thousands of websites received a notification with recommendations to stop using SEO-link and other prohibited ways of website promotion.
May 15, 2015 — First wave of Minusinsk algorithm
According to the announcement, as a result of Minusinsk first iteration, 488 websites rankings have dropped. Yandex continues to encourage all webmasters to abandon the use of SEO-link.
May 27, 2015 — Second wave of Minusinsk algorithm
Second iteration of the Minusinsk algorithm has touched about 200 websites.
Jun 23, 2015 — Third wave of Minusinsk algorithm
Third iteration of the Minusinsk algorithm against SEO-links.
Sep 8, 2015 — Update of AGS
Update of AGS algorithm, which aims to limit the visibility of websites publishing SEO links on their pages. According to the Yandex statement to get out from sanctions, it is sufficient to abandon the placement of SEO links on your website.
Feb 2, 2016 — Vladivostok Algorithm
Change in algorithm of ranking in mobile search results. Yandex begins to consider optimization for mobile devices. Now the search results on desktop and mobile devices may vary. The priority in mobile search results get pages adapted for mobile devices.
Nov 2, 2016 — Palekh Algorithm
Another algorithm that uses machine learning. Palekh is aimed at improving how Yandex handle uncertain long-tail queries, which do not have any behavioral data. Using neural networks it helps to understand the meaning of query and find relevant documents, based on their titles.
Mar 23, 2017 — Baden-Baden Algorithm
The algorithm reduces the ranking of documents containing over-optimized texts.
Apr 7, 2017 — Update of Baden-Baden
Now this algorithm can knock down the entire domain, even if only certain pages contain over-optimized texts.
Aug 22, 2017 — Korolev Algorithm
Korolev is an update of Yandex ranking formula. The algorithm, using neural networks compares the intent of the query and the content of the document, so it allows Yandex to respond more accurately to complex queries. Unlike the Palekh algorithm from 2016, Korolev analyzes the content of the whole page, not just its title.
Aug 22, 2018 — Introduction of the website quality index and removal of the TIC indicator
Yandex decided to move away from the TIC (Thematic Citation Index), which was based on a qualitative assessment of inbound links and replace it with the website quality index (IKS in Russian). When calculating the new index, Yandex uses data about the website and the business that this website represents. These data can be obtained from search results as well as from other Yandex services: Metrica, Maps, Zen and so on.