How to tell if a site sells links

A write for us page I read recently carried five requirements. One dofollow link permitted in the body. Nothing gambling, adult, crypto or payday. Eight hundred words minimum. No links to direct competitors. A reply within twenty four hours.

Five clauses, and none of them about a reader.

That page is a rate card with the price lifted out. The price arrives in the reply.

Two questions, and everybody asks one

The first gets written about. You are considering a placement and want to know what you are buying into.

The second gets ignored, and it is the harder half. A link appears in your referring domains from a domain you have never contacted. No seller to email, nobody who wants your business. You want to know whether that domain is one a search engine already has a file on, and the only material available is the site itself.

This is not seller vetting

I have written separately about vetting a seller before you pay, and that is a different exercise. It is a procedure run inside a conversation: the numbers you pull on their domain, the terms you get into an email. Every step assumes somebody who wants the sale to close.

Nothing here assumes that. These are footprints on a site’s own pages, readable by anyone, including on properties that would never sell you a thing. Half the domains you need to judge are ones you never approached.

If you are buying an entire site instead of a slot on one, this read is one line item inside a much longer diligence list. Separate subject.

Discount me accordingly

I sell placements. A dofollow slot on a comparison article on one of my own properties went for $250 this year on a twelve month term.

So I am describing my own category, and every tell below is one somebody could aim at a page of mine. I think that makes the list sharper rather than softer, because I know which of these I can defend and which would make me squirm.

Outbound first, always

The single strongest signal is what a site links out to. Not what links in. What it points at.

Pull every outbound link, drop the internal ones, group by destination domain, sort by count, read the top of the list.

A site linking out editorially points at sources. Documentation, a study, a news story, occasionally a competitor.

A page factory points at commercial homepages. Not articles on those domains. Homepages. An immigration lawyer in one country, a furniture retailer in another, a sportsbook in a third, an invoicing app in a fourth.

The geography is the part I weight hardest. A real publication builds relationships inside the market it serves. Commercial dofollow links landing in eleven countries across four industries were chosen for one reason, and that reason arrived by invoice.

Outbound is also the hardest thing here to dress up, because nobody bothers. I have never once had a seller volunteer their outbound profile.

I say that as somebody it has embarrassed. A crawl of my own portfolio came back with 15,321 outbound dofollow links against 16 inbound. Almost all of it was template junk, a footer credit and a plugin. But a crawler cannot tell a sold link from a widget, and on somebody else’s site neither can you. Read the number as a reason to look closer.

The archive is a payment log

Get the post dates from an archive page, a feed, a sitemap. Sort them, ignore the volume, look only at the gaps.

Somebody writing about a subject they care about leaves an uneven but continuous line. Weekly for a while, a quiet month, then back.

A site that publishes on payment leaves clusters. Nineteen posts in one week, nothing for two months, fourteen posts in nine days. Orders arrive in batches and get fulfilled in batches, and the archive keeps the receipt.

Lay the clusters against an ordinary business calendar too. Dead through late December, busy in the second week of January, is a publishing schedule behaving like an invoicing schedule.

Cheap to run, and my second most trusted after outbound.

Bylines that appear once

Open the author archive. Most platforms expose one whether the site links to it or not.

A real publication has a few names carrying most of the output and a tail of contributors with two or three pieces each. A page factory has forty names with one post apiece. The byline was made for the article and the article was made for the link.

A genuine contributor’s bio names a job and an employer. The manufactured ones say the author is passionate about technology, over a photograph that turns up on three other sites.

Now the caveat, because this tell is weaker than it looks: legitimate sites run contributor programmes and end up with long tails of people who wrote once and vanished. I have that on my own sites.

The rate card with the price removed

Back to that write for us page, because it is the tell closest to a confession.

Real editorial guidelines talk about the reader. What our audience already knows, the angles we have covered to death, what we will not run, how to pitch. Sales requirements talk about the link. Position, count, attribute, prohibited verticals, minimum length.

The tell inside the tell is the turnaround promise. Editorial teams do not commit to reading an unsolicited draft within a day. Sales teams commit to replying within a day.

Topical scatter, ranked last

Personal finance, dog grooming and online casinos inside the same four weeks. That is a warehouse with categories bolted on the front.

Everybody leads with this one, which is why I put it fifth. It throws the most false positives on the list. General interest blogs exist. And the operators worth worrying about fixed this years ago: the good page factories are single niche and sail through a topical scatter check.

Count the tells, do not chase them

None of the five proves anything alone.

Legitimate sites take guest contributions. Legitimate sites accumulate a strange outbound profile from a plugin installed in 2019 that nobody has looked at since.

So the question is never whether a tell is present. It is how many stack on one domain, and whether each has a boring explanation you can actually go and find. One tell is noise. Two buys the site another ten minutes. Four or five together on a domain with no visible audience and you have your answer.

If you pay for backlink data, a vendor spam score joins that stack as one more input, never as the verdict. It describes a neighbourhood more than a site, argued elsewhere.

When the link found you

Run the same five checks on new referring domains you never asked for.

Clean result, do nothing, which is correct for nearly all of them. Page factory result, still mostly do nothing. The reflex is a disavow file and the reflex is usually wrong.

I learned that slowly. Fifty flagged domains off my own profile, opened one by one, expecting an attack. I found scraper mirrors and sites republishing a feed with the links left intact. The site had simply been indexed a long time.

The inbound check earns its keep on the delta rather than the level. One page factory a month is weather. Thirty inside a fortnight, all landing on the same page with the same anchor, is an event.

The line I would defend

Whether the site would have published the piece anyway.

On my own properties that is one question and every enquiry goes through it. Would this article exist here if nobody had paid. If the topic is one I already cover, the piece is one I would have commissioned, and the link points at something a reader on that page would plausibly want, then money changing hands has not altered the artefact.

A page factory fails at the first clause, because the article would not exist. The article is packaging. The link is the product, and everything around it is there to make the delivery resemble publishing.

That distinction survives the obvious objection, that both parties took money. They did. So does a magazine running an advertorial, and nobody calls a magazine a link farm over it. The question is whether there is a publication underneath the transaction.

You can test it from outside. Find three articles on the site carrying no commercial link at all and read them properly. Competent and on subject, there is something real under there. Cannot find three, there is not.

That test has cost me money. I have turned down enquiries that would have paid because the piece did not belong on the site, and the buyer went and got it somewhere else the same week. That somewhere else is the thing this whole list is about.

What none of this can tell you

It does not tell you what a search engine thinks. I am inferring a category from surface footprints. Nobody outside those companies sees the internal classification, and anyone selling a tool that claims to read it is selling a guess with a logo on it.

I also cannot show you the counterfactual. I have never bought two placements on sites identical in every respect except these tells and watched what each did. Rankings move for a dozen reasons at once and the sample size is one buyer with one budget.

So this is a way of forming a view in twenty minutes instead of an afternoon, from somebody at both ends of the transaction. Judgement with evidence attached.

The one thing I will say flatly is the direction of the error. People who run this check buy fewer bad placements than people who read an authority score and wire the money. The crawl I run and the exact outbound grouping are here.