How to vet a publisher before you place a link
Vet a publisher on relevance, real organic traffic, what it ranks for, its own link profile and how freely it sells links. This page gives the criteria, a 100-point scorecard and a five-minute link farm check.
In short
- Topical relevance is the primary factor for 58.6% of senior SEOs when they judge a link, according to uSERP's September 2026 survey of 116 respondents.
- Only 1.37% of roughly 257,000 sites selling guest posts met a basic quality bar in BuzzStream's 2026 analysis, and 39.6% had no organic traffic.
- Our 100-point scorecard is a proposed working template. It weights relevance at 20 points and traffic in the target country at 15.
- Some findings reject a site whatever its score, including zero organic traffic, cloaking and a seller who will not name the site.
A link is only as good as the page it sits on. Vetting means checking, before you pay or pitch, that the publisher is a real site with a real audience that Google still trusts. The checks below take a few minutes per site and remove most of what is on sale. That filter matters because most of the supply is poor. BuzzStream’s 2026 analysis of roughly 257,000 sites selling guest posts found 96.2% were low quality, 39.6% had no organic traffic, and only 1.37% reached a basic bar of DR 65 or more with 10,000 monthly organic visits. For link insertions, 97.5% of offers were low quality. With figures like those, the starting assumption about any site on offer has to be that it will fail, and the job of vetting is to find the few that do not.
The twelve criteria
The first three criteria ask what the site is about and whether anyone finds it. Relevance comes first: the topic matches at domain, section and page level, and language and country match your market. Second is organic traffic and trend. The site ranks for branded and non-branded terms, gets traffic from your target country, and shows a stable or rising trend over 12 to 24 months with no cliff after a core or spam update. Third is what it ranks for. The keywords match the stated topic. A tech blog that ranks for casino bonus, CBD and payday loan terms is a link farm. These three belong together because each one closes a gap the others leave. A site can look relevant and have no visitors, or have visitors who arrive for subjects that have nothing to do with what it claims to cover.
The next three look at the site’s past and at the links around it. Domain history is the fourth criterion: the archive shows one continuous site, with no switch of topic or language and no expired domain put to a new use. The fifth is the publisher’s own links, where growth looks natural and is not powered by redirects or private networks. The sixth is outbound links. Few posts carry commercial exact-match links, and unrelated articles do not link to casino, adult, pharma or loan sites. The outbound check is the most revealing of the three, because it shows what the publisher has already agreed to do for other buyers, and whatever it did for them it will do on the page that carries your link.
Criteria seven to nine are about whether people edit the site. Editorial signals come seventh: named authors with real profiles, an editorial policy, a list of prohibited topics, and evidence that the site rejects pitches. A site that turns work down is exercising a judgement, and that judgement is what gives a link from it any meaning. Eighth is the publishing pattern, where the volume fits a site edited by people. Ninth is the “write for us” page. A public page with a price, categories covering everything, a generic name, a default theme and an author called “admin” all point the same way, towards a site that exists to sell space.
The last three concern how the placement itself is sold and delivered. Seller-list exposure is the tenth criterion: the domain is not in circulated price spreadsheets, on several marketplaces, or in cold emails offering five sites at once. Eleventh is the technical side. Your article will be indexable, in the sitemap and linked internally, and not parked in a noindexed “sponsored” folder. Twelfth is the deal terms, which cover permanence, link attribute, labelling, replacement, and no competitor link in the same article. A site can pass the first ten and still deliver nothing if the article is hidden from search engines, which is why the technical check is made on the publisher’s existing sponsored posts before you commit and not on your own after it has gone live.
Thresholds practitioners use
These figures come from a guide Setproduct published in September 2026. They are rules of thumb from one source, not measured limits. They are useful all the same, because several of the twelve criteria ask whether something “looks natural”, and a rough figure gives two people a shared starting point for that judgement.
| Check | Normal | Red flag |
|---|---|---|
| Traffic estimates across Ahrefs, Semrush and Similarweb | Spread of about 2x | Spread above 5x |
| Referring domain growth | 20% to 30% a year, with bumps | 200 to 3,000 new referring domains in six months, then flat or a crash |
| Outbound anchors | Mostly editorial | 60% or more exact-match commercial anchors |
| Posts published | 3 to 10 a week on a small edited site | 5 to 15 a day across unrelated categories |
| Offer | Site named, price discussed | Under $100, site name hidden until you reply |
The first row works because the tools estimate traffic from different data, so some disagreement between them is expected and a very large one suggests the visits are of a kind that only one tool is counting. The second row describes a shape as much as a number. Steady growth with bumps is what a site collecting links over the years looks like, while a sudden climb followed by a flat line or a crash is what a bought or redirected profile looks like once the buying stops. The publishing row is a matter of simple capacity, since a small site with real editors cannot turn out posts across unrelated categories at the rate in the right-hand column.
How to spot a link farm in five minutes
The twelve criteria describe a full review. Most sites never need one, because a short run of checks removes them first. The steps are ordered so that the quickest and most decisive come at the start.
- Read three recent posts. Would anyone subscribe to this?
- Check the author. Is there a real person behind the name, with a LinkedIn profile?
- Open the site in Ahrefs or Semrush. Look at the organic traffic trend, the top pages, the top keywords and the shape of the referring domains graph.
- Search Google for the site name restricted to its own domain, together with words such as casino, CBD, payday or escort. Off-topic sold content shows up at once.
- Open five random posts. Count the commercial outbound links and the exact-match anchors.
- Look for “write for us”, “sponsored post” or “advertise” pages with a price list.
- Check whether the site appears in Google News or has any brand search demand.
- Ask for a GA4 or Search Console screenshot with the address bar visible.
If the site fails steps 1 to 4, stop. You do not need the rest. Those four cover reading the content, checking the author, looking at the traffic and searching for sold content, which between them test the things a link farm finds hardest to fake at the same time. The last step is the only one that depends on the seller’s cooperation, and the address bar is asked for so that the screenshot can be tied to the domain being sold. A refusal is information in itself, since a publisher with real traffic loses nothing by showing it.
The 100-point scorecard
This is our proposed template. It turns the criteria into a number so that two people vetting the same site reach the same decision. The weights are a working method, and you should adjust them to your own risk policy. The value of a scorecard lies less in the exact points than in the discipline: every site is asked the same questions in the same order, and a decision to approve can be explained afterwards by pointing at the rows.
| # | Criterion | Points | Scoring guide |
|---|---|---|---|
| 1 | Topical relevance, domain and page | 20 | 20 exact niche, 12 adjacent, 5 general, 0 unrelated |
| 2 | Organic traffic in the target country | 15 | 15 for 10,000 or more a month, 10 for 1,000 to 10,000, 5 for 100 to 1,000, 0 under 100 |
| 3 | Traffic trend over 24 months | 10 | 10 stable or growing, 5 volatile, 0 collapsed |
| 4 | Keywords match the site’s topic | 10 | 10 clean, 5 some drift, 0 ranks for unrelated money terms |
| 5 | Outbound link hygiene | 10 | 10 rare commercial links, 5 mixed, 0 every post sells |
| 6 | Editorial quality and real authors | 10 | 10 named experts and standards, 5 acceptable, 0 “admin” or AI filler |
| 7 | Publisher’s link profile | 10 | 10 natural, 5 unclear, 0 hockey stick or network powered |
| 8 | Domain history | 5 | 5 continuous, 0 repurposed expired domain |
| 9 | Seller-list exposure | 5 | 5 not publicly selling, 3 sells discreetly, 0 on every list |
| 10 | Technical and deal terms | 5 | 5 indexable, internally linked and permanent, 0 noindex, orphan or rental |
What the score means is a matter of four bands. A site that scores 80 or more is approved. A score of 65 to 79 means approve at a capped price and volume, so the site can be used without becoming a large part of the profile. A score of 50 to 64 means use only for second-tier links or diversification, and anything under 50 is rejected. The middle bands exist because vetting produces many sites that are acceptable without being good, and a plain pass or fail would force a choice between wasting them and treating them as equal to the best.
Some faults cannot be outweighed by points elsewhere, so reject the site whatever it scores if any of the following is true. It has zero organic traffic. It is deindexed or shows signs of a manual action. The placement would be hacked or injected into the site without the owner’s knowledge, which should end the conversation at once. The page is cloaked, so Google sees something different from visitors. It links out to unrelated restricted niches, unless that is your niche. The seller withholds the site name, or the seller refuses to show a sample URL. The last two are about the seller and not the site, and they are hard fails because every other check on this page depends on being able to look at the thing you are buying.
Why authority scores are not enough
Domain Rating and its rivals are third-party estimates, and Google does not use them. They can be inflated by redirecting expired domains, by links from high-DR profile pages and by link farms that link to each other. Traffic can be faked too, with rankings for junk or misspelled keywords and visits from irrelevant countries. That is why the scorecard gives most of its points to relevance, traffic in your country and what the site ranks for. Each of those is harder to counterfeit than a score, and they are harder still to counterfeit together, because junk rankings show up in the keyword check and foreign visits show up in the country check.
BuzzStream’s 2026 insertion data points the same way: a site’s organic traffic predicted its price better than its DR did. Put another way, prices in that data followed traffic more closely than they followed the authority score. A buyer who filters by DR alone is therefore relying on the weaker of the two signals in that data, and is exposed to every inflation method listed above.
Once a site has been vetted, record the score for every site in your tracker, then write the placement using how to write a placement brief. Keeping the scores means the next person does not have to vet the same domain again, and it lets you look back later at whether the sites that scored well were the ones whose links lasted. To set the pass mark and the hard fails as company rules, see link buying policy and governance. If you buy through a platform, link marketplaces explains what their filters do and do not check, which matters because a marketplace filter can cover part of the twelve criteria and leave the rest to you.
Common questions
What is a good Domain Rating for a link?
There is no good number on its own. Domain Rating is an Ahrefs estimate that ignores traffic and spam, and it can be inflated with redirected expired domains. Check relevance and real organic traffic first.
How do I know if a site is a link farm?
Look for a public price list, categories that cover every topic, posts by "admin", and rankings for casino, CBD or loan terms on a site about something else. Most link farms fail within five minutes of checking.
How much organic traffic should a publisher have?
Our scorecard, which is a working template, gives full marks at 10,000 or more monthly organic visits from your target country and no marks under 100. Zero organic traffic is a hard fail.
Should I trust traffic numbers from one SEO tool?
No. Compare two tools. A spread of about two times between tools is normal, and a spread above five times suggests the traffic is manipulated, according to a 2026 guide from Setproduct.

