An analysis of multi-tier link-building practices: how link tiers work, why the “one-to-ten” scheme is based not on mathematics, how to calculate a safe acquisition pace, how to measure the impact of an individual link, and where to look for the cause when a site drops after links are acquired. This material is for those who already buy links and want to understand what they are paying for.
Three categories of sites that search engines work with
A useful working model that explains why one link can lift a page while a hundred similar ones do nothing. Broadly speaking, all sites fall into three groups based on how the search engine treats them.
- Premium group. Domains twenty years old or more, expert content, professional forums—medical, engineering, and industry-specific. They rank in the top three for almost any query of their own, and a link from one of these sites works at full strength.
- The middle tier. Around ten years old, with decent traffic, but flaws in the content or technical setup. These sites tend to rank around positions three to ten. Links from them work, but are two to three times weaker.
- The gray mass. Everything languishing on the seventh to tenth pages of the results: new domains the search engine has not classified yet, copied content, and AI-generated text without editing. Such links have no effect and can sometimes do harm.
Even within the premium group, there is a nuance that eats up half of many budgets: you need to get into a legitimate section. A link from a category where the site has been accumulating paid articles on every subject for years—and especially one from a subdomain—will perform like a link from the gray mass, even if the domain is considered exemplary.
This is also where the assessment of links in Yandex comes from. If we take the effect of a quality link in Google as one hundred percent, the same link in Yandex delivers roughly thirty to forty percent. The difference is not that Yandex ignores links, but that it takes far fewer of them into account.
The gray mass is the only category where Google and Yandex differ fundamentally. In Google, it produces an effect at industrial volumes: one and a half to three million links consolidated through tiers onto a single page. In Yandex, such volumes do not work at all, so nobody builds multi-tier structures for it.
Why DR is a poor guide when buying links
A situation that should raise suspicion: the domain is two or three years old, yet its rating is close to seventy. There are few possibilities—the site was boosted with redirects from purchased expired domains, or it was deliberately built up for the metric. The expired domains are hidden from SEO services and left visible only to search bots, so you will not see the backlink profile.
The story repeats itself. Ten years ago, the Runet was manipulating TCI in exactly the same way: the higher the metric, the more expensive the links sold for, and a site with a TCI of 2000 supported its owner better than a Moscow salary. Now DR is being inflated on an industrial scale, and people have been doing this for five to seven years already.
A practical sign: if a seller starts the conversation with the rating rather than traffic and topical relevance, the site was most likely created to sell links. You should look at other things:
- the domain’s history in the web archive—whether the site’s profile changed and whether it is a restored expired domain;
- traffic dynamics over several years: stable or growing;
- how the site weathered updates over the past two or three years.
When choosing between a site with growing traffic and one with high metrics, take the first one without hesitation. Growing traffic under current conditions means that professionals are working on the site and doing it correctly. Such sites are worth monitoring: once a month, check which links have appeared, what formats they have started using, and where the new traffic came from. It is inexpensive reconnaissance and a source of ideas at the same time.
Link tiers: how the pyramid works
The terminology is simple. The first tier consists of pages that link to your site. The second, third, and subsequent tiers consist of links that boost those pages. The point of the structure is to strengthen existing placements rather than create new direct links.
The first-tier rule
Link-building software—GSA Search Engine Ranker, Khrumer, RankerX—is never directed at your own site. Runs are aimed only at the first tier.
Ideally, you should control the first tier. The best option is your own satellite sites or blogs on free platforms such as Blogspot and WordPress.com. The reason is straightforward: if a first-tier link is removed, the entire pyramid built beneath it disappears into thin air along with the budget. That cannot happen with your own site—you can always change the link or redirect it to another project.
The “one-to-ten” ratio and its real reason
The scheme looks like this: for every link on the previous tier, there are ten links on the next one. One press release, ten links to it, a hundred to those ten, and a thousand on the next tier.
This ratio is held together not by mathematics, but by human relationships. Place a hundred thousand software-generated links on someone else’s paid article, and the site owner will write to say they do not need spam and will remove the article or replace it with a redirect. Nobody will notice ten links: in ordinary life, it simply looks as though someone linked to the article. The methodology’s author says that in all his years of using this approach, he has never had a conflict with a site owner, and he can comfortably scale ten links further into the hundreds and thousands.
Schemes for specific situations
If twenty links point to a site, there are several workable options:
- Place five second-tier links on each of the twenty, then another five on each of those. The budget grows with every tier, so the ratio is chosen according to the available funds.
- Place one publication from an inexpensive regional news site on each link—not a capital-city outlet or a million-visitor site, but a second-tier publication—and strengthen the news site itself with three to five links. This works well for narrow niches where there are few placement opportunities.
- Split the twenty links into ten pairs and boost each pair according to its own scheme. After two to four weeks, compare which group produced growth.
Two tiers are usually not enough for a commercial site if both are built with free software. But if the second tier consists of paid placements on news sites that get into news aggregators, and they are additionally strengthened with five to ten rented links, two tiers are quite sufficient.
Tiers are not built with software alone. A completely manual structure looks like this: a press release linking to the site, two or three paid articles linking to it, then notes from your own blog and contextual mentions in signatures on relevant forums linking to those articles. In the Runet, tiers can be assembled entirely from rented marketplace links, with the budget distributed in descending order: the second tier is more expensive, the third is mid-priced, and the fourth is the cheapest, used simply to speed up indexing.
What to place on each tier
The closer a tier is to your site, the more important link quality is. The closer tiers need contextual links—links surrounded by text: articles, notes, and wiki-engine content. Comments, social bookmarks, gallery signatures, and other minor links go on the distant tiers, where all that is required of them is quantity and indexing.
A separate note on scale. The methodology’s author record is roughly ten million links accumulated on one project over several years, but that was an experiment, not standard practice. In real work, the ceiling is one and a half to two million links on the first tier; beyond that, scaling further is pointless because there are simply no new referring domains left in the databases.
Speeding up indexing with redirect links
An underestimated type of link is a page warning that the user is about to proceed to an external site. Many large sites show an intermediate page when an external link is clicked, such as “you are leaving for another site,” with the target page’s address contained directly in the intermediate page’s URL. Google indexes such pages readily.
They are collected by scraping footprints—characteristic parts of URLs—and iterating through the letters of the alphabet to expand the sample. Then one and a half to two million such links are layered on top of already placed links to improve their indexing.
They do not bring a site to the top by themselves: this was tested separately. Their function is to diversify the profile and speed up inclusion in the index. An interesting detail: similar redirect links also appear on Google’s own subdomains in different countries, and they are indexed even better than links from ordinary sites.
Incidentally, there is no universal indexing-acceleration service—you have to keep several and alternate between them.
The principle of maximum diversity
The principle underlying the entire approach is: “a system cannot detect something that has no system.”
This is why the approach rejects any fixed ratios of anchor and non-anchor links. Instead of a single formula, a ratio table is created—90 to 10, 80 to 20, 60 to 40, 50 to 50, and so on—and the link builder randomly selects a ratio from it for a given period. If many links are being placed, the period is a day; if there are two or three per week, it may be a week or a month. Ratios may repeat consecutively or not be used at all. No pattern emerges in the end, so there is nothing to detect.
The same approach is used to distribute links among the homepage, categories, and internal pages, as well as to boost satellite sites. Fifty satellite sites run through with the same service package form a ready-made pattern that is easy for an algorithm to spot.
The reason for rejecting popular recommendations such as “no more than twenty percent anchor links” is observation of sites with genuinely natural backlink profiles. We are talking about enthusiast blogs: for example, an Argentine stamp collector who has been publishing catalog-based analyses of issues since 2011 and has no idea what SEO is. Such sites build communities around themselves, and links come from equally passionate people; there are no ratios at all: sometimes they have one hundred percent non-anchor links, and sometimes the reverse. All the percentages that circulate in recommendations were derived by someone from their own sample and declared to be a rule.
There are still two exceptions to the lack of system. First, a link from an expensive, high-trust site should be an anchor link—paying serious money for it and using a bare URL is wasteful. Second, the younger the site, the greater the share of non-anchor links in its profile, and it is better to preserve this imbalance.
Non-anchor links and the text around them
The surrounding text is crucial for a non-anchor link. A bare URL dropped into a message without explanation looks like spam to both people and algorithms. The minimum is a short phrase explaining what is behind the link: the address, a dash, and a two- or three-word explanation containing keywords.
Search engines take the words near a link into account—both Google and Yandex. This gives a non-anchor link topical relevance without formally making it an anchor link.
With anchor links, the surrounding text is secondary, but there is still one rule: do not repeat in it what is already in the anchor. If the anchor contains a key phrase, use a synonymous formulation in the explanation; otherwise, you create needless keyword stuffing.
Non-anchor links also work in internal linking. The only requirement is that page URLs be short. A long URL made from a hundred-character title wraps across several lines on a mobile screen and harms both appearance and usability.
How to calculate a safe acquisition pace
The methodology replaces guesswork about the “naturalness” of a particular number with concrete figures.
- Take ten to twenty competitors and select about ten healthy ones: five to seven years old or more, with stable or growing traffic, a sustained presence in the top results, and no drops after updates.
- For each one, record the growth dynamics of referring domains: the month when the first links appeared, how many domains there were after six months, after a year, and in subsequent years. In shorthand: 10 → 50 → 100.
- Put everything into a table—sites horizontally, checkpoints vertically—and calculate the average. Remove sites that stand out sharply from the rest from the calculation: they experienced a link spike due to viral content, not systematic work.
- Evaluate your own site not by its calendar age but by its link age. If it already has some link mass, it is not a newborn site but a six-month-old or one-year-old site.
- Stick to the established pace, laying out the plan immediately for two or three years ahead.
With this approach, the way a link is acquired is secondary. A marketplace, outreach, or manual placement—it makes no difference, as long as the donor is topically relevant and healthy. More details on selecting donors and the signs that mean a site should be rejected immediately can be found in a separate analysis of the selection methodology.
How to measure the effect of an individual link
Before placement, only a forecast is possible: the site's traffic trends, topical relevance, the indexing speed of new content, and the expected number of visits. There is no tool that can say, “this link will work with eighty percent certainty,” and there probably never will be one.
The effect is assessed in practice after placement, over a period of two to four weeks. Sometimes a link works within a few hours; sometimes the result takes more than a month to appear, but the average interval is precisely that.
You need to look at three things at once:
- rankings for the target query cluster, not just for a single keyword;
- traffic to the specific page—the link can have an effect even without visits, by improving rankings;
- search visibility: how many queries the page ranks for in the top-50 and top-100.
A condition without which measurement loses its meaning: do not dump fifty links onto one page. Spread the placements across different pages; otherwise, you will not be able to tell which link caused growth and which one dragged the page down. The same approach answers the question “does the link hurt”: if rankings for the cluster fall after placement, the cause is most likely the link.
When a site drops after link acquisition begins
A real-world case: two sites dropped after link acquisition began, one tenfold and the other threefold. The contractor was experienced, and the donors had been checked. This usually means that a certain share of the donors turned out to be toxic, and specifically according to parameters that standard checks do not examine.
Where to look:
- the number of outgoing links from the donor and their ratio to incoming links;
- the metrics of the sites the donor links to—it may be part of someone else’s spam network;
- the domain's presence in spammer databases and malware databases;
- signs of hacking. A telling case: a domain was hacked for spam distribution from its mail server; no links or doorway pages were left behind, and this was discovered only through checks against spam databases;
- the trend in the number of the donor's pages in the index. Growth from 2500 to 3000 over a year is normal. A jump to twelve thousand means the site was sold, doorway pages were uploaded, or it moved to a new CMS and service pages were accidentally left open—much like attachment pages in WordPress used to be indexed.
This leads to a useful habit: when placing a link, record the donor's key metrics, including the number of pages in the index, so that six months later you have something to compare against. It is best to remove toxic links by agreement with the site owner; anything that cannot be removed should be submitted to disavow—based on the methodology author's observations, the tool is processed after all, even if not immediately.
Economics: PBNs, Wikipedia, and finding contractors
Your own network. PBNs make sense where they save money. This includes expensive niches such as crypto and real estate, where placements immediately cost twice as much, or regions with a shortage of sites—Scandinavian countries and small European markets. The second argument is control: around ten to fifteen percent of “permanent” links disappear within a year. The site is sold, the owner starts squeezing it for advertising without moderation, or the site turns into a dump—and the link stops working even though it is formally still in place.
Wikipedia. The methodology's author gave up on it: the links do not repay either the time or the money. A link from a strong page that receives traffic and has many external links is a different matter, but obtaining one is more difficult, takes longer, and costs more. If you have such a link, great; if not, there is no reason to be upset.
Journalist services. HARO, renamed Connectively, provided links from sites that were otherwise inaccessible, at reasonable prices. The service finally shut down on 9 December 2024—Cision folded it into its main platform. There are alternatives on the market, but the era of cheap access to journalists is over.
Price intelligence. Services such as Link Detective and seorestore.com maintain downloaded marketplace databases and show the minimum price for a specific domain. Here is one way to save money: an owner lists a site on several marketplaces at once; sales go better on one of them, so the owner raises the price there and forgets about the second. The difference between marketplaces for the very same domain can be severalfold.
Finding contractors. Links costing a dollar apiece in competitive niches deliver nothing: the service exists solely to make money for the contractor. Link-building offers on LinkedIn are almost entirely spam from contractors who undercut prices to reach clients with budgets. Freelance marketplaces are no better: most packages are useless, though not harmful.
A practical testing approach is to launch a test site or take one site with a dozen pages on different topics and promote each page with a separate service. Then compare the results by rankings, traffic, visits, and visibility. On average, one worthwhile contractor emerges from every ten to twenty orders, and that is the person worth working with long-term.
Geography, language, and domains
Links from sites in the language of the project being promoted work even if the sites themselves are in another country. A Lithuanian site can be strengthened perfectly well with links from Lithuanian resources serving the diaspora in the United States—the language is the same, so everything is fine. A link in another language, however, will perform noticeably worse.
The Baltics nevertheless behave as a single cluster: the languages are similar, and Google treats the region accordingly. A broader observation, according to an author acquaintance who heads a group of Russian-language linguists at Google's New York office, is that within the company languages are grouped by language families, and the citation of content is counted within those groups. Judging by practice, links work the same way.
When entering a new region, the search results themselves answer the question of which domain zone to choose. If national domains predominate in the country's commercial top results, you need to register a national domain, including dealing with local business-registration requirements. If the top results are occupied by domains in international zones, entry is easier—and this matters in places where domains are issued only to residents who provide a passport.
Southeast Asia and the Arab region are currently undervalued, in the author's assessment. The main obstacle there is not technical but cultural: without understanding the mentality, a site is built blindly. An illustration from the US market: design is secondary for an American audience, as Craigslist, Yelp, and similar sites show—the important thing is whether the page answers the question, not how attractive it is. The Russian approach of meticulously polished landing pages does not transfer well there.
Small things that change the result
A link is born together with the page. When it is inserted into an article published a year ago, search engines see this and interpret it unambiguously: it was added artificially, most likely for money. For this reason, marketplace placements “in existing articles” have almost no effect.
Outgoing links from your site. The problem is not the number as such but the type of page. A list with one hundred and fifty links to services or one hundred and twenty links to plugins in a single article looks like a spam page. A reasonable guideline is to have fewer outgoing than incoming links. Four hundred outgoing domains against one thousand five hundred incoming links is a normal ratio.
Checking a news site through an aggregator. Before buying a placement, copy the site's domain and search for it in the search engine's news tab. If the latest indexed news appeared half an hour or two hours ago, the news bot visits regularly, and your publication will enter the index quickly. If the latest update was a week ago, there is no need to hurry.
A format for building brand visibility. A paid article presented as an interview with the company's founder or leading specialist is perceived differently from an ordinary sponsored publication: with an introduction, questions, photographs, and thanks at the end. The author reached this conclusion while analyzing a leak of Google's internal documentation and confirmed it through experiments. Press releases presented as news—regional and industry-specific—fit the same logic.
Internal linking. It still works, but external links are more important. Start with external links; work on internal linking in parallel. A pure silo structure is almost never found on live projects, more often appearing as a partial implementation; semantic cocoons—one parent article and a dozen child articles on subtopics—make it possible to gather authority where it is needed right now.
A competitor with over-optimization in the top results. If a site with eleven percent exact-match anchor occurrences remains at the top, it is not a model to copy. Something else pulled it to the top, while the over-optimization simply has not reached the threshold at which penalties begin. You should compare yourself not with one such site but with the average values across a dozen or two competitors, discarding anomalies.
Sequence of actions
- Select a dozen healthy competitors and calculate their average growth rate in referring domains—this is your safe corridor for the next year or two.
- Divide existing links into those you control and those that depend on someone else's goodwill. It makes sense to build pyramids around the former.
- For each first-tier link, plan a reinforcement scheme at a ratio of about one to ten and record it so you can compare options later.
- Build the near tiers from article and wiki placements with text around the link; build the distant tiers from anything that gets indexed.
- Create a table of the proportions of anchor and non-anchor links and select values from it randomly for each period.
- Spread placements across different pages and record rankings, traffic, and visibility two to four weeks after each placement.
- Record donor metrics at the time of purchase, including the number of pages in the index, so you can later track the site's degradation.
- Test new contractors on a separate site rather than the live one, and do not expect results before the tenth order.
