SEO in Yandex in 2026: what changed and what no longer works

Three structural shifts that changed SEO in 2026

The first shift is that Yandex now indexes mobile-first, without exceptions. The primary index is built from the mobile version of the page, so anything hidden, collapsed or stripped out for the sake of a lighter phone layout barely exists as far as search is concerned. The classic mistake is showing specifications, prices, reviews and tables on desktop while burying them on mobile inside an accordion that loads over JavaScript after a tap. The desktop version used to act as a safety net. It does not any more: if the price is not in the mobile HTML, search concludes the page has no price.

The second shift is that algorithms score the document as a whole instead of summing isolated factors. Neural ranking models compare the page against the job the user came to do, not against a checklist of keyword occurrences and meta tags. The old logic of adding another 2,000 characters and three more keyword mentions stopped producing results. A page either solves the problem, with prices, terms, timelines and answers to objections, or it does not, and keyword density will not compensate. Individual factors no longer stack linearly: a flawless technical audit on a page with nothing to say buys you nothing.

The third shift is answers delivered inside the results page. Yandex Neuro and Search with Alice, along with Google AI Overviews, resolve informational queries without sending the user anywhere. Impressions in Webmaster keep climbing while clicks stay flat, and that is the normal picture in 2026 rather than a broken analytics setup. The practical conclusion is straightforward: informational traffic is losing value while commercial, branded and long-tail queries with clear intent are gaining it.

Together these shifts change where the budget should go. Money that used to buy word count and links is better spent in 2026 on mobile performance, on the completeness of commercial information, and on making the company visible beyond its own domain through maps listings, reviews and mentions.

• Open a service page on a phone and check whether price, phone number and the order button are visible without expanding anything.
• Compare impressions against clicks in Yandex Webmaster over 12 months. A widening gap is the zero-click effect, not a tracking bug.
• Check which pages Webmaster flags as low-value or low-demand. That label is the algorithm telling you what it thinks of the document.
• Confirm that mobile and desktop serve the same content rather than a trimmed-down phone edition.

What stopped working, stated plainly

Templated pages spun across dozens of queries. The pattern of one master text plus two hundred city and service substitutions worked while ranking depended on word matching. Today the model recognises the template by its structure, not by a uniqueness percentage, and hundreds of near-identical pages end up flagged as low-value, dragging down the assessment of the whole site. If only five of your hundred regional pages have a reason to exist, keep five and fill them with real data: addresses, regional pricing, local cases, branch contacts.

Bulk link buying from marketplaces. Link-based ranking in Yandex has been written off: the influence of purchased links on commercial results is close to zero, and recent algorithm updates only reinforced that. Meanwhile a sudden spike in marketplace backlinks is still an easily readable manipulation signal. In budget terms that is zero return at best and exposure at worst. Links matter in Yandex in exactly one form, as a natural side effect of PR: a piece on vc.ru, coverage in an industry publication, a listing in a relevant directory.

Keyword stuffing. A H1 reading like a search query, the key phrase in every other paragraph, a wall of anchor text in the footer, all of it points straight at the Baden-Baden text filter. In 2026 that filter reacts to unnatural phrasing rather than a formal density threshold: if a sentence does not read like something a person would say, it works against the page. The test is simple. Read the paragraph out loud.

Behaviour manipulation. This is where most agency blogs turn vague, because they still sell the service. Plainly: in 2026 faking behavioural signals is not a grey but workable tactic, it is an asymmetric risk. Anti-fraud systems identify emulation through session, device and network patterns, and the penalty is not a gentle drop of a few positions but effective removal from results for months, with a slow and uncertain recovery. Betting a business that lives on organic traffic against that is economically absurd: you save a few months of honest work and risk the entire channel.

• Stop: cloning regional and service pages, buying link packages, appending SEO filler at the bottom of pages, ordering behaviour manipulation.
• Start: cutting the page count and raising the density of useful information on the pages that remain.
• Open Pages in Search and then Excluded in Webmaster to see how many documents the algorithm has already discarded.
• If a contractor builds behaviour manipulation into position-based KPIs, that is a reason to end the contract rather than negotiate.

E-E-A-T in Russian practice: show experience instead of claiming it

Yandex does not use the term E-E-A-T. That wording comes from the Google quality rater guidelines. The signals behind it, though, are proxied by Yandex through its own mechanisms: author pages, organisation data, reviews on Maps, branded search demand, and how well the content matches the declared specialisation of the site. Working on expertise pays off in both engines, but the tasks need to be stated more concretely than raise our trust score.

The scarcest of the four letters is first-hand experience. Most articles in the Russian-language web are a retelling of three other articles, and models distinguish that better every year. Experience shows up in details a compilation never has: concrete numbers from practice, screenshots of the actual interface, a description of what went wrong, honestly stated limits of the method, a walk-through of an edge case. One paragraph with real data outweighs a page of generalities.

The second signal in order of importance is organisational transparency. A company that shows no legal details, registered address or contract terms is indistinguishable to an algorithm from a fly-by-night site. For YMYL topics such as medicine, finance, legal services and construction, licences, permits and verified author credentials get added on top. The checklist below covers both layers.

• An author page: name, role, years of practice, relevant education or certification, links to industry publications or a professional profile.
• A byline with publication and update dates, where the update actually changes the content rather than just the date.
• An about page with tax ID, registration number, legal address, bank details and the public offer contract.
• Licences, professional permits and verified specialist credentials for YMYL topics.
• Original photographs of the office, team, production and completed work instead of stock imagery.
• Numbers, timelines, limitations and edge cases drawn from your own practice, inside the text itself.
• Links to primary sources: documents, official Yandex help sections, industry research with the publisher named.

Behavioural factors: what genuinely improves them

Behavioural factors are not the time-on-site number in your analytics. Search evaluates the whole chain of query, results page, site, return. The key signal is whether the user came back to the results and clicked a competitor next. If the session ends on your site, the job was done. If the person returns and picks the neighbour, the page lost regardless of how many minutes it held attention.

What actually moves those metrics. A first screen that matches the wording of the query, so the visitor recognises what they searched for within two seconds. A direct answer in the first three to five lines instead of after three paragraphs of throat-clearing. Speed: LCP under 2.5 seconds on mobile and no layout shifts. No pop-ups covering content above the fold. And completeness, so nobody has to go back to search for the rest of the answer.

What does not help despite looking sensible. Splitting material across paginated screens to inflate pageviews, hiding key information behind a click, telling people to read to the end. All of it raises formal engagement numbers while lowering real satisfaction: the visitor spends longer hunting for the answer, gets annoyed and goes back to the results. Optimise for time to answer, not time on page.

A separate behavioural signal is snippet click-through. A title that echoes the query wording and carries specifics, a description with a price or a timeline, sitelinks, rating and price pulled from structured data all lift CTR at an unchanged position. Rising CTR at equal footing improves the position by itself, so rewriting the top 30 titles is often cheaper and faster than producing new content.

Commercial ranking factors: 60 to 70 percent of the weight

With link-based ranking written off, commercial factors took over as the main differentiator in Yandex commercial results. Industry estimates put them at roughly 60 to 70 percent of the ranking weight for commercial queries. That does not make text irrelevant: text decides whether the page enters the relevant pool at all. Inside that pool, though, commercial signals set the order.

The logic matches common sense: when two pages answer the query equally well, the one that makes buying easier goes higher. Easier means a visible price, an obvious next step, a phone number and legal details, described delivery and warranty terms rather than a line telling the visitor to ask a manager.

One nuance matters: commercial factors are assessed both at page level and at site level. A single landing page with a price will not save you if the site has no contacts, no legal details and no delivery section. That makes them easiest to fix as a package, in one sprint against the table below. This is the fastest-return work in SEO, since it needs neither links nor months of waiting for a full recrawl.

Off-site trust signals: reviews, maps and brand mentions

The Yandex Business listing on Yandex Maps is the most underrated asset in Russian SEO. It feeds local results, contributes to overall domain trust and frequently generates more enquiries than the website, because people call straight from the card. Treat it like a page: complete categories, real photographs, current hours, a price list, replies to every review, regular posts. A half-filled listing with a 3.8 rating and three reviews from two years ago costs the business more than it looks.

Reviews outside your own site. Beyond Maps that means 2GIS, Zoon, Otzovik, IRecommend, vertical platforms such as Profi and Avito Services, and marketplaces for product businesses. Algorithms care about the freshness of the stream and the presence of company replies, not just the average score. Ten reviews from the last quarter outweigh a hundred from three years ago. Collecting them has to become a process: an email or message after each closed deal, a QR code on documents, a reminder from the account manager.

Unlinked brand mentions. Articles on vc.ru and Habr, expert commentary in trade media, appearances in roundups and rankings, conference talks: together they build what search treats as company recognition. A separate and very strong signal is branded demand, the growth of searches containing your company name in Wordstat. It is hard to fake cheaply and it correlates directly with algorithmic trust.

A practical habit: once a quarter, check what the first page of results shows for your company name. If nothing but the website appears, no listing, no reviews, no coverage, then you are competing purely on the content of your own domain, which is the most fragile position there is.

• Bring the Yandex Business listing to full completion: categories, photos, hours, price list, description, links.
• Set up a review request after every closed deal and reply to all of them, negative ones included.
• List the company in three to five vertical directories and aggregators relevant to your niche.
• Plan one or two external publications per quarter: vc.ru, Habr, trade media.
• Track branded search demand in Wordstat as a standalone metric in your reporting.

The zero-click reality and the bridge to GEO

The 2026 reality is that a large share of informational queries is resolved inside the search interface itself. Search with Alice, Neuro, AI Overviews, quick answers and rich blocks hand the user a condensed response and the click becomes optional. For content projects that undermines the traffic-equals-revenue model. For a service business it is more a change of metric: what matters is no longer the visit count but whether the company gets named inside the answer.

What to do about it in practice. First, stop judging SEO by traffic alone: add impression share, branded demand and leads to the report. Second, shift content effort toward queries AI does not resolve, meaning specific commercial queries, local intent, comparisons with prices, configurators and calculations. Third, restructure material so the brand ends up inside the generated answer.

That last point is GEO, optimisation for generative engines. One common misconception is worth killing here: Yandex neural search does not maintain a separate index, it operates on top of the regular one. GEO is an add-on to SEO, not a replacement for it. Without baseline technical quality, relevance and domain trust, no amount of optimising for neural networks produces anything. How that layer works and how to measure it is covered in our separate article on GEO.

A 90-day priority plan

Sequence matters more than completeness. The most common mistake is starting with content: articles get written while the site stays slow, priceless and full of duplicates. The right order follows return on effort: technical foundation and commercial factors first, content and expertise second, off-site signals third.

Month one is the foundation. A technical audit of the mobile version, speed work, correct robots.txt, sitemap.xml and canonicals, removal of duplicates and junk pages, prices and contacts on every commercial page, a legal details block, and pages for delivery, payment and warranty. By the end of the month the site should satisfy the entire commercial factor table.

Month two is substance and trust. Rework the 10 to 15 priority pages around the real user task rather than keyword occurrences; add author pages and bylines; implement Organization, Service or Product and FAQPage markup; clear out low-value pages using Webmaster data; rewrite titles and descriptions for the top 30 pages by impressions.

Month three is the external perimeter and the GEO layer. Yandex Business listing, a review collection process, registration in vertical directories, the first external publication, a branded demand baseline. In parallel, prepare pages for citation in generative answers: direct answers under headings, tables, factual density. Position movement is usually visible by this point and traffic starts to compound.

• Days 1 to 30: technical base, speed, duplicates, commercial factors across all service and product pages.
• Days 31 to 60: rework priority pages, authorship and E-E-A-T, structured data, low-value page cleanup.
• Days 61 to 90: Yandex Business, reviews, directories, external publications, preparation for AI citation.
• Continuously: weekly rank tracking, monthly query review in Webmaster and Metrica, with leads as the priority metric rather than positions.

Frequently asked questions

Related services

• SEO and GEO website promotionSearch traffic is the one acquisition channel that keeps working after the ad budget stops. We run two tracks in parallel: classic SEO for Yandex and Google, and GEO, the work of getting your pages quoted inside AI Overviews, Yandex Neuro, ChatGPT and Perplexity, where a growing share of queries now ends. It starts with an audit, never with buying links.
• GEO: getting your brand cited by AI searchMore and more queries end in an AI answer rather than a click. GEO is the work of getting named in that answer, with a measured mention share rather than a promise.
• Website development, end to endWe build sites that still hold up a year later: from a one-page CMS build to a store or a web application on React and Next.js. One team for design, code, integrations and launch.

Read next

• GEO in 2026: how to get cited by ChatGPT, Alice and AI OverviewsGEO is not a separate channel, it is a layer on top of SEO. Here is how six answer engines pick sources, what makes a page quotable, the honest truth about llms.txt, and a weekly method for measuring AI visibility.
• Yandex and Google: how promotion differs and how to work with bothOne site, two search engines, two different ranking logics. Where the differences actually change the work, why regional settings matter so much in Yandex, and how to build a strategy without writing everything twice.
• A Slow Site Loses Money: Core Web Vitals and ConversionSite speed is not a technical metric, it is a line in the revenue report. Core Web Vitals explained in plain language, how to find the real cause of a slow page, and a fix list sorted by return per hour of work.