Link exchanges and reciprocal links

By Bill Hartzer · Hartzer Consulting · 25+ years in search, domains and expert witness work

What a link exchange is

A link exchange is an agreement: I put a link to your site on mine, you put a link to my site on yours. It is the oldest deliberate link building tactic there is, it predates the phrase link building, and it is the only tactic in this reference that Google's own documentation describes in something close to plain speech.

Google's link spam policies list, among their examples of link schemes, excessive link exchanges — rendered in the documentation with the parenthetical link to me and I'll link to you — and partner pages that exist purely for the sake of cross-linking. That phrasing has been in Google's guidelines in one form or another for the better part of twenty years. There is no ambiguity to interpret, no dispute among practitioners about what the policy covers, and no argument that the tactic has been misunderstood. It is named.

The modern forms are dressed up but structurally identical. Two site owners agree to place links in body content rather than on a links page. Three owners route it as A links to B, B links to C, C links to A, so no pair links directly. Groups on messaging platforms coordinate hundreds of participants with shared spreadsheets. Agencies run partner networks among their own clients. Guest post swaps — I publish your article, you publish mine — are a link exchange with an article wrapped round it.

All of it is the same trade, and the trade is what the policy prohibits.

How the scheme works, step by step

Described honestly, so you can recognize it when it is offered to you.

1. Find a partner with comparable metrics. The matching is usually done on Domain Rating or Domain Authority, because participants want an even trade. This matching is itself a detectable pattern — natural links do not cluster by metric similarity between reciprocal pairs.
2. Agree the placement. Historically a links or partners page. Now more often a contextual link inside an existing article, because participants know footer and links-page swaps are obvious.
3. Agree the anchors. Each side specifies the phrase they want. This is the step that converts a mutual recommendation into a transaction, and it is the step that leaves the clearest evidence.
4. Place both links, usually within days of each other. Temporal clustering is the second clearest piece of evidence.
5. Scale it. One exchange is nothing. The tactic only produces measurable results at volume, so participants join groups, build lists, and run dozens or hundreds of trades. Volume is what makes the pattern statistically undeniable.
6. Attempt to hide it. Three-way and four-way routing, link wheels, staggered placement dates, varied anchors. Every one of these increases coordination cost without removing the underlying structure.

What it costs in time and effort

The nominal cost is close to zero, which is the entire appeal. No money changes hands in a pure swap, no content has to be created, and a competent operator can arrange dozens of trades in a week. That cheapness is why the tactic refuses to die.

The real costs are elsewhere and they are large. Every reciprocal link you accept means placing an outbound link on your own site to a page you did not choose on merit, which degrades your own pages for your own readers. Coordination overhead grows fast once you move to three-way routing, because you are now maintaining a graph rather than a list. And the recovery cost, if the profile is ever the subject of a manual action or a broad devaluation, is enormous: identifying, contacting and removing hundreds of reciprocal links is months of work, and disavowing them does not undo the outbound links you placed.

Priced properly, a link exchange costs you an outbound link, a coordination burden, and a contingent liability. The zero on the invoice is the least informative number in the transaction.

When it works and when it does not

It does not work, and the reason is mathematical rather than editorial. A search engine holds the web as a graph of directed edges. A reciprocal pair is a bidirectional edge. Counting bidirectional edges is one of the cheapest computations available on a graph — far cheaper than judging content quality, evaluating relevance, or anything else a ranking system does routinely. Computing the share of a domain's inbound links that are reciprocated is trivial, and comparing that share against the distribution for sites in the same sector is trivial too.

Three-way routing does not solve this. A cycle of length three is only marginally harder to detect than a cycle of length two, and the operators who run these schemes leave far louder signals than the cycle itself: the same participants appearing together across many cycles, near-simultaneous placement, matched anchor optimization, and links inserted into articles where they make no editorial sense.

Now the part that gets missed. Two sites linking to each other is not, by itself, a problem. The web is full of unremarkable mutual links: a manufacturer and its distributor, two co-authors citing each other, a conference and its speakers, a supplier and a customer case study, two open-source projects that depend on each other, a parent company and its subsidiary. These happen constantly on ordinary sites and Google's own policy language addresses excessive exchanges and pages built purely for cross-linking, not the existence of reciprocity.

The distinguishing question is simple: was the link the reason, or the consequence? If you would link to them regardless of whether they linked back, it is an ordinary link. If your link exists because theirs does, it is an exchange. A useful test is to imagine removing their link to you. If you would then remove yours, you already know what you built.

Common mistakes

• Believing three-way exchanges are undetectable. They are the tactic's founding myth. Cycle detection in a graph is a solved problem and has been since long before search engines existed.
• Calling it a partnership. Renaming the trade does not change its structure, and the partners page is the most recognizable artifact in the history of link building.
• Swapping guest posts. An article each way with a link each way is an exchange with more steps, and it produces two footprints instead of one.
• Joining exchange groups. A shared spreadsheet of participants is a ready-made map of the scheme, and it exists on a platform you do not control.
• Damaging your own pages. Every accepted swap puts a link you did not choose in front of your readers. Do this fifty times and your site is a directory of people who did you a favor.
• Assuming reciprocity is inherently bad. The opposite mistake. Refusing to link to a genuinely useful site because they happen to link to you is superstition, and it makes your site worse.
• Not measuring your own reciprocal rate. Most people running exchanges have never calculated the number that would tell them how exposed they are.

A worked example

Consider two sites in the same market, and look at what an analyst sees.

Site A runs exchanges. Over four months it acquires 180 new referring domains. Of those, 168 also receive an outbound link from Site A. In 141 cases the two links were placed within nine days of each other. The anchors pointing in are commercial phrases; the anchors pointing out are commercial phrases too. Twenty-two of the partners are themselves linked to each other. Forty of the inbound links sit in articles on unrelated topics, inserted mid-paragraph where a reference makes no sense. There is a page titled Our partners with sixty entries.

Site B has 180 new referring domains over the same period. Nine of them also receive links from Site B: a component supplier whose product it documents, an industry body it belongs to, two conferences it spoke at, three customers it published case studies about, and two tools it recommends in its documentation. The placement dates are spread across months with no relationship to each other. The anchors are brand names and titles. Nobody linked because somebody else did.

Site B has a five percent reciprocal rate and it means nothing. Site A has a ninety-three percent reciprocal rate with matched timing and matched anchors. No human judgment is needed to tell them apart — the second one falls out of a query. That is why this tactic is graded high risk while ordinary mutual linking is not graded at all.

How to measure it

If you have inherited a profile and want to know your exposure, these are the numbers to compute.

• Reciprocal rate. The share of referring domains that also receive an outbound link from you. Most link tools will export both sides; the intersection is a spreadsheet operation. Single-digit percentages are ordinary. Anything approaching half needs an explanation.
• Placement date proximity. For each reciprocal pair, the gap between the two links appearing. Natural pairs are scattered across months and years. Coordinated pairs cluster within days.
• Anchor symmetry. Whether both directions use optimized commercial anchors. Genuine mutual links almost always use brand names, titles or bare URLs on at least one side.
• Editorial fit of the inbound placement. Read the host paragraph. If the link has been dropped into a sentence that did not need it, that is the tell a human reviewer will use, and it is the one that survives every attempt at obfuscation.
• Cluster overlap. Whether your reciprocal partners link to each other. Exchange groups produce dense clusters; the ordinary web does not.
• Your partners page. If one exists, count it and read it honestly. That page is the single most incriminating artifact on most sites that have run this tactic.

The verdict

Avoid it. This is not a marginal call and I am not going to soften it.

Excessive link exchange is named in Google's link spam policies in language a beginner can understand. It is detectable by the cheapest computation available on a graph. It costs you outbound links from your own pages, which is a real and permanent price paid in the quality of your own site. The routing tricks invented to conceal it have never worked and add coordination overhead that makes the scheme harder to run than legitimate outreach would be. And the recovery, if it comes to that, is longer and more miserable than anything you gained.

None of that means you should be nervous about the ordinary case. Link to the supplier whose product you use, the conference you spoke at, the tool you recommend, the research you cite. Some of them will link to you. That is what an interlinked professional field looks like and it has never been a problem.

The line is the trade. The moment the link becomes a payment for another link, you are in the policy, and the fact that no money moved is not a defence anyone has ever successfully made.