Playbooks
Outsourcing Link Building in 2026: What to Buy, What to Never Buy
Most link building gets outsourced, and most outsourced link building is bad. The difference is almost never the vendor's pitch deck. It is whether you scoped the work around something a machine can verify, or around a number in a spreadsheet.
Link building is the most outsourced job in SEO and the one that goes wrong most often. Not because agencies are dishonest, though some are. Because the thing being bought is almost always specified wrong.
A typical scope reads: twenty links per month, DR 50 or above, dofollow, niche relevant. Every term in that sentence is measurable, which is what makes it feel like a good spec. None of them describe whether the link will do anything.
Here is what actually changed in 2026, what to buy, and how to tell within thirty days whether you hired the right people.
The compensation model is the whole problem
Pay someone per link and you have told them exactly what to maximize. They will maximize it. That is not a character flaw, it is the incentive you designed.
The cheapest way to produce a link that satisfies “DR 50, dofollow, niche relevant” is a site built to satisfy that description and nothing else. These sites exist in enormous numbers. They have a plausible name, a few hundred posts on scattered topics, a Domain Rating propped up by links from other sites in the same network, and a monthly traffic figure somewhere near zero. They will happily publish your guest post in four days for $180.
You will get your twenty links. Your rankings will not move. When you ask why, the report will show the links were delivered as specified, and that will be true.
The fix is not a better vendor. It is a spec that cannot be satisfied cheaply.
What to buy instead
Scope link building around placements, not links. A placement means: a specific page, on a specific site, that a specific audience reads, containing a reference to you that a person would follow.
Concretely, replace the usual metrics with these.
Organic traffic to the linking page, not the domain. Domain-level authority scores are trivially inflated. Page-level traffic is not. A page that gets 400 organic visits a month is worth more than a DR 70 domain whose linking page gets none. Ask for the traffic estimate of the exact URL before approving.
Editorial resistance. If a site accepts every pitch, the link is worth what it costs, which is nothing. Ask your vendor what their rejection rate is on outreach. A healthy answer is somewhere between 85 and 97 percent rejected. If they tell you 40 percent of pitches convert, they are pitching sites that sell links.
Topical proximity that survives a sentence test. Write the sentence a reader would see: “For a deeper look at inventory forecasting, see this guide from X.” If that sentence would look strange on the page, the relevance is fake regardless of what category the site is filed under.
A named human on the other end. Placements arranged with an editor, a writer, or a founder tend to survive. Placements arranged with an anonymous “content team” email tend to disappear in eight months when the site is sold or cleaned up.
What changed now that AI engines read the web
Links used to have one job: pass ranking signal to Google. They now have two, and the second one is changing how the first should be scoped.
Answer engines decide which sources to quote partly by how often a source is referenced elsewhere and by whom. A mention without a hyperlink still counts here. Being named in a roundup, quoted in a trade publication, or cited in someone’s research all contribute to whether ChatGPT or Perplexity treats you as a reasonable thing to cite, even when no link exists.
This has three practical consequences.
First, unlinked mentions stopped being consolation prizes. If your vendor negotiates a mention in a well-read industry piece but cannot get a hyperlink, that is still a result. Count it. Vendors paid per link will not report it, which tells you something about the compensation model again.
Second, the sites worth targeting shifted. Publications that answer engines already quote heavily are worth more than their Domain Rating suggests. You can find them: run the ten prompts your buyers would run, note which domains get cited in the answers, and hand that list to your vendor as the target set. It is the single most useful hour you can spend on outreach strategy, and almost nobody does it.
Third, the link-farm problem got worse, not better. A site with no real readers is not cited by anything, so a link from it contributes nothing to your citation profile either. The two systems now agree on which links are worthless, which is convenient.
Vendor models, ranked by how often they work
Retained agency, full service. They prospect, pitch, write, and place. Costs $2,000 to $10,000 a month. Works when you have volume needs and no internal capacity. Fails when you stop reviewing the target list, which happens around month four in most engagements.
Marketplace or database purchase. You browse a catalog of sites, pick, and pay per placement. Fast, transparent pricing, and almost every site in these catalogs is there because it sells links. Occasionally useful for a specific niche publication. Not a strategy.
Digital PR agency. They pitch stories, not links. Costs more, converts less often, and the placements that land are genuinely good: real publications, real readers, real citation value. Works when you have something worth pitching, meaning original data, a strong point of view, or an actual event. Fails when you hand them a product page and hope.
Freelance outreach specialist. One person, $1,500 to $4,000 a month, usually 3 to 8 placements. The quality range here is enormous and depends entirely on the individual. When it works it is the best value in the category, because you are buying one person’s relationships rather than a process.
In-house with outsourced prospecting. You keep relationships and approval, someone else builds the target lists and does first-touch outreach. Underrated. Prospecting is the part that scales badly with seniority and the part that most agencies do worst.
Real pricing, without the hedging
Editorial placements on sites with meaningful traffic: $250 to $1,500 each in ordinary niches. Finance, legal, health, and gambling start at $2,000 and go up steeply, because the sites know what the traffic is worth.
Digital PR campaigns: $5,000 to $25,000 per campaign, producing anywhere from 2 to 40 placements depending on whether the story lands. High variance is the point of the model.
Agency retainers: $2,000 to $10,000 monthly. Below $2,000, the math forces the vendor toward cheap placements. Do the arithmetic yourself. If a retainer promises fifteen links at $1,500 a month, each link costs the agency under $100 including labor. You already know what kind of site that buys.
Anyone quoting DR 60 placements at $50 is selling network links. Not maybe. There is no legitimate cost structure that produces that number.
The thirty-day test
You do not need six months to know whether an engagement is working. You need the first batch of placements and about twenty minutes.
Take every URL delivered in month one and check four things.
-
Does the linking page get organic traffic? Any traffic tool will tell you. If the answer is zero across most of the batch, stop the engagement.
-
Who else does that page link out to? Open the page and look at the other outbound links. If it links to a casino, a CBD store, and a SaaS product in the same post, you are on a link farm and so is your brand.
-
Would you have wanted this placement if links did not exist? The most reliable question in the category. Ask it about each URL. The honest answer is usually immediate.
-
Did the vendor report any unlinked mentions? If they placed twelve links and zero mentions, either they are not pitching real publications or they are not counting the outcomes that do not fit their invoice.
Three failures out of four means the spec was wrong. Rewrite it around placements and traffic before you renew.
What to keep in-house, always
Two things, and they are small.
The target list. You decide which sites are acceptable. The vendor can propose, and should, because prospecting is their skill. But approval stays with you, in writing, before outreach starts. This single control eliminates most of the failure modes in this article.
The claim about your own business. Whatever a guest post says about what you do, what you sell, and what you have achieved, you approve the wording. Outsourced writers invent credentials. Not maliciously, they are filling a gap in a brief. It ends up on the permanent web with your name on it.
Everything else can go. Prospecting, first-touch outreach, follow-up sequences, drafting, editorial back-and-forth, tracking. None of it needs to be yours.
The honest summary
Outsourcing link building works. It works when you buy placements on sites that people read, pay a price that makes that possible, keep the veto on targets, and check the first batch properly instead of reading the summary row of a report.
It fails when you buy a number.
The AI shift did not change that. It just added a second scoreboard that happens to agree with the first one, and made the sites worth targeting slightly easier to identify, because you can now see which ones the engines already trust.