Buying links at scale: the anchor-text pitfall that caps the returns

A quick note on sourcing, because it matters. This is an anonymized audit of a real, large paid link program across a portfolio of brands. I have stripped every identifying detail, generalized the figures to ranges, and named no company, vendor, or site. What is left is the pattern, which is the part worth learning from, and it is a pattern I see constantly.

And this is not a callout of anyone. A program like this is a rational answer to what clients ask for: predictable, paced monthly volume. The people who run these programs work hard, deliver real placements on real sites, and have revenue targets to hit, just like the rest of us. The pitfall I want to flag is structural, baked into how paid placement works, not a knock on anyone’s effort.

The program, by the numbers (anonymized)

The scale was serious: several thousand live links built across a brand portfolio, at roughly a few hundred dollars a link, adding up to a low-six-figure annual spend. My first surprise was that this was not a junk network. The host sites were legitimate: a median domain rating in the low 60s, real five-figure monthly traffic per referring page, only a small fraction under DR 40. Someone was paying for quality inventory.

The composition is where it gets interesting. Essentially every link was a placed, paid type: about half guest posts, a third link insertions and niche edits, the rest link exchanges. Editorially earned links, the kind a writer adds because your page deserves the citation, were statistically nonexistent in the deliverable. This is a pure buy, on good sites.

The tell: an inverted anchor profile

If you only get to look at one number on a backlink profile, look at the anchor text distribution. Across years of analyses, a natural profile lands in a fairly consistent shape: roughly 30 to 50 percent branded anchors, 15 to 25 percent naked URLs, 15 to 25 percent generic phrases like “read more,” 10 to 20 percent partial match, and only about 3 to 8 percent exact-match commercial anchors.[2]Industry anchor-text analysesNatural anchor text distributionIndependent analyses converge on a natural profile dominated by branded anchors and naked URLs, with exact-match commercial anchors a small single-digit share. Above ~10 to 15 percent exact match is the over-optimization danger zone.View source ↗ Real people, linking for real reasons, mostly use your brand name or your URL.

The program I audited was the mirror image: roughly four times as many commercial, keyword-rich anchors as branded ones. That is not what happens when humans cite you. It is what happens when someone is buying anchors to move specific keywords. And it is self-defeating, because that exact shape is the fingerprint the algorithm looks for.

One important caveat, because it is the difference between reading a deliverable and reading a domain. That commercial-heavy number is the anchor profile of the built links, not of the brand’s whole backlink graph. On a large, established site with tens of thousands of referring domains, a few thousand commercial anchors barely move the overall ratio, so the domain-level profile can still look perfectly natural and the over-optimization risk is lower than the subset implies. The inversion matters in two narrower ways: as a detectable pattern within the placed-link set itself, and at the domain level for smaller brands, where bought links are a large share of the total. Read the number as the shape of what was bought, not the shape of the site.

What Google actually counts, and how it knows

The common objection is fair: Google cannot see the email thread or the invoice, so how can it know a link was paid? The answer is that it does not need to. It infers paid links from patterns, and it has three good ones here.

• The anchor ratio itself. An inverted, commercial-heavy distribution is a statistical signal that links were engineered for ranking. Google’s spam policy names “optimized anchor text in articles, guest posts, or press releases” as a link scheme by example.[3]Google Search CentralSpam policies: link spamOptimized anchor text in guest posts and articles is explicitly listed as an example of a link scheme. The anchor profile is a first-class signal, not an afterthought.View source ↗
• Site-level signals. Google crawls the same advertising, “write for us,” and contributor pages that sell placements. A site that systematically sells links can have its outbound link equity discounted at the site level, while its own genuinely useful pages keep ranking. Your instinct that Google can flag the seller and quietly stop trusting its outbound links is essentially how this works.
• Footprints. The same referring domains reused across a set of related brands is a classic cross-site footprint. Patterns that repeat across a portfolio are easier to catch, not harder.

Guest posts are a special case, because Google has been explicit. John Mueller has said for years that links inside guest posts you place should be nofollow, that guest-post links are frequently unnatural, and that Google has trained its systems on years of data to catch and devalue them algorithmically.[4]Search Engine JournalGoogle’s warning on guest posting for linksMueller: guest posting for links results in unnatural links; links you place in guest posts should be nofollow; Google has years of data to catch and devalue them algorithmically, mostly without manual review.View source ↗ A program that is half guest posts is half-built on the tactic Google has spent the longest learning to ignore.

How much of the equity is likely neutralized

Since 2022, Google’s link-spam system, SpamBrain, has been pointed specifically at detecting links bought and sold to pass credit, and at nullifying that credit rather than penalizing the site.[5]Google Search CentralLink spam update, now powered by SpamBrainGoogle: SpamBrain detects both sites buying links and sites selling them, and neutralizes the link credit. The link stays up; the equity stops flowing; no notice is sent.View source ↗ A program that is entirely paid, majority guest posts, commercial-anchor-heavy, and footprinted across sibling brands is squarely in that crosshair. My estimate for this profile is that well over half of the intended ranking equity is being neutralized, concentrated in the commercial-anchor guest posts.

The saving grace, and the durable version of the spend

Here is the part that keeps a program like this from being a total write-off, and it is also the lesson. Because the host sites are real and high-traffic, the placements still deliver two things even when the ranking equity is discounted: referral traffic, and brand mentions on credible sites. And branded mentions increasingly drive visibility in AI answers, where they correlate with citations far more strongly than backlinks do.[6]Ahrefs · 75K brandsWhat correlates with AI Overview visibilityAcross roughly 75,000 brands, branded web mentions correlate with AI-answer visibility about three times more strongly than backlinks. The mention can matter even where the link is devalued.View source ↗ So the money is not entirely burned. It is just buying the wrong thing the expensive way.

Which points straight at the better version. If real-traffic placements with branded framing are the part that survives, then engineer for that on purpose: earn placements on relevant, real-traffic sites, lead with branded and natural anchors instead of commercial ones, and target the pages and topics that actually move the business. You get fewer links per dollar than a guest-post mill, but a far higher share of them are durable, and you are building brand and AI visibility at the same time. The audit’s real takeaway is not “paid links are dead.” It is that the only part of paid links still working is the part that looks earned, so you may as well earn it.

“But isn’t paying someone to build links the same thing?”

Fair question, and I get it a lot, especially because I run a link-building service myself. If you pay a vendor for links and you also pay me to build them, what is the real difference? It comes down to who gets paid, and for what.

Paid link acquisition pays the publisher. Money goes to the site owner in exchange for placing your link. The link exists because money changed hands, and that is the transaction Google’s guidelines define as a link scheme, no matter how good the host site is.

White-hat link building pays for labor, not placement. The money pays for the work: building an asset worth citing, finding the right pages, and the outreach to convince a publisher to add the link editorially, for free, because it genuinely helps their readers. The publisher is never paid. The link exists because the asset deserved it.

It is the same reason an in-house SEO’s salary, or a digital-PR agency’s fee, is not “buying links.” You are paying for the work of earning coverage, not for the coverage itself. Journalists are not paid; the team that pitched them is. The honest test is one question: would the link survive if Google knew exactly how it happened? A link a publisher chose to add, with no money flowing to them, passes. A placement you paid them to insert does not.