← Back to all articles
strategy

Semrush Counts 307 Sites Linking to Us, Google Shows 19: the Spam Network Padding Agency Backlink Counts

Last reviewed:

Semrush says 307 websites link to conectia.pro. Google's Search Console shows 19.

I found the gap between 14 and 16 November 2026, while benchmarking the link profiles of Spanish software agencies in Semrush to plan our own link work. Most of the difference is not our doing. We did not buy those links, and we are not the only target: the same automated spam network has multiplied the referring-domain counts of every agency profile I opened in detail before publishing, mostly between September and November 2026. I opened 25 more later that day, and their figures are in the update further down.

A disclosure before the numbers. Conectia competes with the sites in this sample for the same searches, which is why none of them is named here. Our own figures are named because they are ours. We don't sell SEO services. We use Semrush on a trial account and have no other relationship with it, or with any link seller.

Four in five of the sites Semrush says link to us arrived after August 2026

Semrush's chart of our referring domains sits at 20 to 30 from December 2025 to July 2026, reaches about 60 in August, and climbs to about 290 by the end of October. About 250 of the 307 arrived from September onwards, and almost none have been lost since.

What arrived is easy to describe:

  • 298 of the 307 referring domains (97%) have an Authority Score of 10 or less. Four are above 20. One is above 40.
  • 69 are .shop domains and 30 are .site.
  • The two most frequent anchor texts are sales templates: «high quality dofollow backlinks da 50 pa 40 premium pbn network service … buy backlinks online cheap» and a fake testimonial about someone's «SEO game».
  • Semrush's own Network Graph flags our profile as «Suspicious».

Search Console tells a different story. Its links report lists 19 sites: directories and review profiles, a Reddit thread, two rankings published by other agencies, our own sister properties, a client's footer credit and three scraper or site-statistics pages. None of them is a .shop, .site or .online domain. Google describes the report as a sample of the links it knows, so 19 is what Google chooses to show me, and I can't read it as a count of everything its crawler has seen. What I can say is that the 250-odd newcomers are absent from it.

The same pattern appears in every profile I opened in detail before publishing

Semrush lets you open a full backlink overview for one site at a time. Besides ours, I opened three Spanish agencies that compete with us for the same searches. The timelines are read by eye from Semrush's charts and are approximate; the other columns are the interface's own figures.

Table: The same pattern appears in every profile I opened in detail before publishing
SiteReferring domains nowBefore the jump (chart)Share with AS 0–10Domains with AS above 20Authority Score over the period
Site A352~175 (Aug 2026)86%39flat at 23 for 12 months
Site B4,557~400–600 (2024 to mid-2026)98%53flat at 22–24
Site C2,587~300 (Aug 2026)98%3714 → 17 (June) → 22 (Oct)
conectia.pro307~60 (Aug 2026)97%46 since May 2026

Across A, B, C and our own site, more than 6,600 of 7,800 referring domains arrived after August 2026. Site B gained up to about 1,276 new referring domains in a single week. A fifth profile, a software company in another market, grew steadily with no jump but shows the same mix: 88% of its referring domains at Authority Score 0–10, .xyz and .shop domains, and anchors that advertise Telegram accounts.

The fingerprints repeat. Semrush warns that many of the backlinks of sites B and C share IP subnets: B's links come from 5,503 IP addresses packed into 346 subnets. The top anchors of site C are «powerful guest…», «expert seo back…» and «telegram @seo_a… link indexing». Site A shares the «high quality do…ks online cheap» template with us.

Semrush's Backlink Gap report makes the network visible from another angle. I compared our domain with four rival sites and sorted the domains that link to them by how many of the four they share. On the first page of 100 rows (of 7,145), 85 domains link to all four rivals at once. 83 of those 85 have an Authority Score of 5 or less. Eight are the same name registered under eight different second-level domains (.it.com, .us.com, .br.com, .de.com, .eu.com, .gb.net, .uk.com, .uk.net).

Across 57 agency sites, the referring-domain count no longer tells you much

The wider table covers 57 sites: the ones that rank for the searches Conectia targets in Spain and the US (custom software, AI consultancy, fractional CTO) plus four Spanish agencies we follow. They are a convenience sample of our own competitive field, not a random sample of the sector. I read them in Semrush on 15 November 2026, and that table gives only totals, not the composition of their links.

Two things stand out:

  • 18 of the 57 sites have an Authority Score of 10 or less, and 7 of those 18 have between 1,839 and 4,401 referring domains. By count they look like established agencies; their Authority Score says otherwise. On 14 November the same anchor templates appeared in the anchor lists of six small Spanish sites, four of them among those seven.
  • The median site in the table has 2,098 referring domains. Sites with 1,000 to 2,500 referring domains have a lower median Authority Score (18) than sites with 500 to 1,000 (22).

If you rank agencies by referring domains today, you are partly ranking them by how much attention this network has paid them.

Update, 16 November: 25 more profiles, and 24 of them have at least 85% of their referring domains at Authority Score 0–10

Later on 16 November, with this post already written, I opened the backlink overview of 25 of the Spanish sites in that table, one by one. I chose them because they already looked affected: an Authority Score of 10 or less, or more than 1,000 referring domains. That makes them a selected sample. They show what an affected profile looks like, and they say nothing about how many agencies are affected.

  • In 24 of the 25, at least 85% of the referring domains have an Authority Score of 0 to 10. The median profile is at 97%, and the range runs from 81% to more than 99%.
  • Added together, 51,967 of their 54,128 referring domains (96%) sit in that band. The median profile has 20 referring domains above Authority Score 20.
  • The seven sites from the list above, the ones with an Authority Score of 5 to 9 and thousands of referring domains, are all in this group. In each of them the share is above 95%.
  • 23 of the 25 carry at least one link-selling template among their five most frequent anchors. 14 show the «high quality do…ks online cheap» template that also tops our own profile, four show a Telegram handle, and in five the pitch already comes in Spanish. Of the two without a template, one shows only its own terms and the other shows anchors for betting sites, a different kind of spam.
  • .shop is one of the three most frequent top-level domains in 24 of the 25, and the most frequent in 12.
  • Semrush prints its shared-subnet warning on 16 of the 25. They are the 16 profiles with more than 1,800 referring domains. None of the nine with fewer than 1,200 has it, and eight of those nine carry the sales anchors, so a small profile can have the network and no warning.

Two limits. For these 25 the extraction kept only the lowest and the highest point of each 12-month chart, so I can't date their step the way I did for the first four profiles. In 22 of the 25 the low is under a quarter of the high, which puts the growth inside the last twelve months and no more precisely than that. The counts are also still moving: a day after the 15 November table, the overview gave more referring domains for 22 of the 25 sites and fewer for none.

Authority Score mostly held, and I can't show the network moved it

My first internal read, in our notes of 15 November, was that the network inflates Authority Score. The data doesn't support that, and I tested it two ways.

The first test is the 57-site table. The rank correlation between referring domains and Authority Score is 0.50 (Spearman, n = 57). That is a real relationship, and it's what you'd expect, because genuine links grow both numbers. It is also weak enough to leave room for the seven sites above. The correlation between Semrush's traffic estimate and Authority Score is 0.83 (n = 51), but traffic is one of Authority Score's inputs, so that figure is partly circular.

The second test is the timelines. Semrush builds Authority Score from three things: link power (quality and quantity of backlinks), organic traffic, and spam factors that check the profile for manipulation. Three of the profiles I can follow month by month held their score while their counts multiplied: site A at 23, site B at 22–24, us at 6. Site C rose from 14 to 22, but the rise started in June, before its jump in August, and the same profile carries a run of links from regional digital press sites, 52 of them tagged sponsored. I can't separate the network's effect on site C from the press links, so I won't claim one.

The defensible claim is narrower: the referring-domain count is noise; Authority Score mostly resisted it.

What the network changes, and what it leaves alone

It changes the numbers people put in slides. Referring domains is the headline figure in most backlink tools, the one agencies report to clients and founders compare against rivals. For small sites in this sample it now measures exposure to the network more than reputation. It can also make a clean site look like a buyer: Semrush called our profile «Suspicious» for links nobody here asked for.

As far as I can see from outside, it doesn't change Google. Search Console doesn't show these domains, and Google wrote in its December 2022 link spam update that SpamBrain detects sites used to pass outgoing links and that the update neutralises the credit those links pass. Pages carrying more than 2,000 referring domains on a single URL still rank on page one for the most competitive queries in our sample, so I see no sign of a penalty either. I can't see inside Google's ranking, and that is as far as the evidence goes.

Why does the network exist? I don't know. My reading, and it is only a reading: the anchors are adverts. «Buy backlinks online cheap» and a Telegram handle say nothing to a search engine and a great deal to whoever opens a backlink report, which is an SEO, a marketer or an agency owner. Spraying links at small agency sites puts the advert in front of exactly the people who buy links.

Seven signs show whether the network has reached your own report

Five minutes in any backlink tool will tell you whether you have it:

  1. A step in the referring-domain chart between September and November 2026, with almost no lost domains afterwards.
  2. 85% or more of referring domains at Authority Score 0–10.
  3. Cheap top-level domains: .shop, .site, .online, .xyz, .space, .cv, .art, .icu.
  4. Anchors that sell links or name a Telegram account («premium seo backlinks», «expert seo backlink services», «tg @…»).
  5. Families of domains that recur across profiles: one name on many second-level domains, «backlink» in the domain, auto-generated «website worth» and statistics pages about your site, web directories on .xyz with no real categories, unrelated-topic shells (tyres, wallpapers, job boards), escort and casino sites.
  6. A shared-subnet warning, which Semrush prints in the overview. In the 25 profiles of the update it appeared only above 1,800 referring domains, so its absence on a small profile rules nothing out.
  7. Absence from Search Console's top linking sites for your own domain.

What I'd do if you see it

  1. Stop benchmarking on raw referring domains. Count the referring domains above Authority Score 20 instead. On that measure the three Spanish rivals have 37, 39 and 53. We have 4. That gap is unflattering, and it's the honest one.
  2. Use Search Console for your own site, and read third-party totals as a record of what one crawler found.
  3. Annotate the jump in any report you send a client or a board, so nobody reads September's spike as a campaign result.
  4. Don't buy from the senders and don't answer the Telegram handles. The anchors are the sales pitch, and buying links for ranking purposes is the first example in Google's spam policies of link spam.
  5. Don't start a disavow hunt. Google's guidance says most sites will not need the disavow tool, and that you should disavow only when you have a considerable number of spammy links and they have caused, or are likely to cause, a manual action. Links you didn't place, with no manual action in Search Console, don't meet that bar. The same page says Google «works very hard to make sure that actions on third-party sites do not negatively affect a website».
  6. Spend the effort on the tail. The rivals' 37 to 53 good domains come from local business bodies, universities, regional press and directories. That work is slow, and the network can't fake it.

Eight figures from the sample

  • Semrush counts 307 referring domains for conectia.pro; Search Console shows 19.
  • 298 of 307 (97%) have an Authority Score of 10 or less.
  • About 250 of 307 (four in five) arrived from September 2026 onwards.
  • In the first four detailed profiles (three rivals and ours), 86–98% of referring domains sit at Authority Score 0–10.
  • In the 25 profiles added on 16 November, a selected sample, 24 have at least 85% of their referring domains at Authority Score 0–10, and the median is 97%.
  • 85 of the top 100 domains in a five-site Backlink Gap link to all four rival sites at once.
  • 7 of the 18 sites with Authority Score 10 or less have 1,839 to 4,401 referring domains.
  • The rank correlation between referring domains and Authority Score across 57 sites is 0.50.

Every figure comes from Semrush's interface and Search Console, read between 14 and 16 November 2026

I read the Semrush data in its web interface on 14, 15 and 16 November 2026, using our own scripts to pull each table from the page; none of it comes from a Semrush export. The timelines were read off Semrush's charts and are approximate. The 57-site table comes from Semrush's bulk backlink analysis (one row per root domain, global index), read on 15 November. The detailed profiles come from Semrush's backlink overview, and there are 30 of them: ours, three rivals and one site from another market, read between 14 and 16 November, plus the 25 of the update, read on 16 November. I chose those 25 from the 57-site table because they already looked affected (an Authority Score of 10 or less, or more than 1,000 referring domains), so they are not a random sample. For them I have each overview's totals, Authority Score bands, three most frequent top-level domains and five most frequent anchors, not the full list of links. The Search Console figure comes from the links export for conectia.pro, 549 source URLs grouped into 19 registrable domains. Correlations and counts were computed with a short script on those files. The overviews of 16 November already give higher counts than the table of 15 November, and all of them will keep moving as Semrush recrawls.

The network costs you nothing in Google, as far as anyone outside Google can tell. It costs you accuracy every time someone ranks agencies by a number in which, for our own site, 97 of every 100 entries have an Authority Score of 10 or less. If you've seen the same step in your report, I'd like to compare notes.

Work with the authors

The CTOs who write this also build

Fractional CTO engagements, AI engineering squads and an AI development platform we install on your codebase in 24 hours. If this piece matched how you think, the conversation is short.