Building a Link Profile That Holds Up

Relevance Before Volume

The single most common mistake in bought link profiles is optimising for the count. Fifty links from fifty general-interest blogs, each with one article on your subject written for the occasion, is a pattern. Ten links from ten sites that already cover your subject, sitting in articles that would exist whether or not you had paid, is a profile. The difference is not the number; it is whether each link makes sense in the place it sits.

Google's SEO starter guide frames links as something other sites give because your content is useful to their readers. A link that could only have been placed by payment fails that test on inspection, and the more of them a profile holds, the easier the inspection. Topical proximity is the first filter: before asking what a site's authority score is, ask whether an editor on that site would plausibly have linked to you unprompted.

This is why the 84 publishers on this directory are grouped into 8 subject hubs rather than sorted by score. The largest, Business, holds 23 of them. A buyer in that subject has 23 candidates where the topical fit is already established and can spend their attention on the numbers that separate them; a buyer in another subject should be looking at a different hub, not at the highest DA on the site.

Anchor Text Is a Footprint

Anchor text is the most visible signal a link carries and the easiest to get wrong. A natural profile is dominated by brand names, bare URLs, and phrases like "this article" or "their guide" — because that is how people link when they are not thinking about search. A profile where a third of the anchors are the exact commercial phrase the site wants to rank for was assembled by someone thinking about search, and it reads that way to anything designed to notice.

Google's spam policies list "links with optimized anchor text in articles or press releases distributed on other sites" as an explicit example of link spam. That is not a warning about anchor text in general. It is a warning about the pattern where the anchor is chosen for ranking rather than for the reader — which, in a bought placement, is the default unless you decide otherwise.

The practical rule: for every placement you buy, write the anchor the way the host site's own editor would have written it. Usually that is your brand name, sometimes a plain description, occasionally a partial match. If you find yourself negotiating for an exact-match anchor, you are negotiating for the footprint.

Velocity, and Why Spikes Get Noticed

Link velocity is the rate at which a domain acquires links over time. A real site's velocity is uneven but continuous — some months more, some fewer, no month zero and no month a hundred. A site that gained no links for two years and then forty in a fortnight has a velocity graph with a cliff in it, and cliffs are what anomaly detection is built to find.

The mechanism is not published and the specifics are inference, so it is worth being precise about what is known. What Google has documented is that it takes action against link schemes and that its systems attempt to identify them; the reasonable reading is that a sudden burst of similar links from similar sources is the kind of thing such systems are tuned to see. Whether a particular spike triggers anything is not something anyone outside Google can say. What can be said is that a spike is visible, that it looks like what a scheme looks like, and that nothing about a legitimate placement requires it.

The rule that follows is simple: pace it. Ten placements over ten months leaves a graph that looks like a site people gradually found. Ten in a week leaves one that looks like an invoice.

A Mixed Profile

A profile made entirely of one placement type is a fingerprint, whatever the type. All guest posts, all niche edits, all directory listings, all from sites in the same authority band — each is a pattern that would not occur if the links had arrived on their own. The strongest bought profile is the one that most resembles an unbought one, and unbought profiles are mixed.

Guest posts and link insertions have different strengths, and the comparison between them is worth reading before choosing either. Beyond those two, the mix should include the links that cost nothing: a listing in the directories your industry actually uses, a citation on a resource page that already lists your competitors, a mention from a partner or supplier who would link to you if asked. These arrive with the anchors and the velocity a natural profile has, and they dilute the bought portion.

Two further diversities matter. Vary the authority of the hosts — a profile where every link comes from a site scoring between DA 45 and DA 55 was bought to a specification, and the specification shows. And vary the page type: homepage links, deep links to specific articles, links to a category page. One target URL for every link is a pattern too.

Auditing What You Already Hold

Before buying anything, export the profile you have. Ahrefs and Semrush both produce a list of referring domains with the anchor text, the target page, and the date each link was first seen. Sort it four ways.

By anchor text, to see what proportion is exact-match commercial phrasing. If it is more than a small minority, the next several placements should use brand and plain anchors until the balance moves. By referring domain, grouped by subject, to see whether the profile is topically coherent or a scatter of unrelated sites. By first-seen date, to see the velocity graph and any cliffs in it. And by host authority, to see whether the hosts cluster in one band or spread across the range.

Then read the hosts themselves the way you would read a site before buying on it. The checks in the vetting post and the post on networks apply to links you already hold just as much as to ones you are considering. A link you bought two years ago from a site that has since become a network is a link from a network.

Links You Regret

Google's guidance on qualifying outbound links is written for the site giving the link, but it tells the buyer something too: a paid link is supposed to carry rel="sponsored", and a host that publishes paid placements without it is the host exposing both parties. When you audit, note which of your bought links carry the attribute. The ones that do are not passing ranking signal — that is the point of the attribute — and were never going to. The ones that do not are the ones with both the benefit and the risk.

For a link you now regret, the options are to ask the host to remove it, to ask them to add the attribute, or to use the disavow tool. The disavow tool is blunt and Google describes it as intended for cases where a site has been the subject of a manual action or expects to be; it is not a routine maintenance tool and should not be reached for first. Removal by the host is cleaner and, on a site that took payment, usually available for the asking.

The better answer is to not need any of this, which means reading the host before paying. The cost of twenty minutes of checking is low. The cost of a manual action is every ranking the links were bought to earn, plus the recovery.

Deciding What to Buy Next

With the audit done, the next placement is a decision about which gap in the profile to fill, not a search for the highest number. If anchors are over-optimised, the next link is a brand anchor. If hosts cluster on one subject and you sell into another, the next link is on the second subject. If every host sits in one authority band, the next is above or below it. If the profile is all guest posts, the next is an insertion into an existing page, or a citation that costs nothing but an email.

Only then filter the candidates by the numbers. On this directory that means the full table, filtered to the subject, sorted by keyword count rather than by authority, with each traffic figure divided by the keywords behind it before it is taken at face value. As of 29 Sep 2026, the median across the 75 listed publishers where that division can be done is 14.0 visits per ranking keyword. A candidate many times above that owes its traffic to a few terms, and the article you place will start with none of them.

Then read the host. Open the homepage, open two articles, open the robots.txt. Ask whether an editor there would have linked to you unprompted, and whether the anchor you have in mind is one they would have written. If both answers are yes, the link will hold. If either is no, the number that attracted you was measuring something else.

Building a Link Profile That Holds Up.