Every guest post marketplace sells you a homepage. You see a domain rating, a traffic estimate, a category, and a screenshot of a site that looks like a publication. What you are buying is a page on that site, and the two are not the same thing.
We screened 3,805 domains from our own shelf this week, using checks that cost nothing and take about twenty seconds per site. One in six failed outright. Here is what they failed on, and how to run the same checks yourself before you pay anyone.
The homepage is the most respectable page on any domain
This is the whole problem in one sentence. A homepage is designed. It is the page the owner cares about. Everything else is where the money is made.
One site we looked at presents as an educational games platform. Clean homepage, plausible category, good metrics. Its sitemap contains ten pages promoting online casinos, several targeting a non-English market, plus pages for cracked mobile apps, and outbound links to two operators. None of that is visible from the front page or from any metric a marketplace displays.
That is not an unusual case. It is the normal case.
Check one: read the sitemap, not the site
Almost every site publishes /sitemap.xml, and most declare it in /robots.txt. It lists every URL the site wants indexed. Scan those URLs for the words that give the game away: casino, slots, betting, poker, matka, escort, viagra, steroids, buy-instagram-followers, mod-apk, essay-writing.
Across 3,805 domains, the flags broke down like this:
| category | domains flagged |
|---|---|
| gambling | 951 |
| adult | 301 |
| essay mills | 191 |
| pharma | 165 |
| follower selling | 122 |
| cracked software | 57 |
Gambling dominates, by a wide margin. If you learn one pattern, learn that one.
Check two: a flagged word is not a verdict
This is where most naive checking goes wrong, and it is the part nobody writes about.
Our first pass flagged an Indian construction trade magazine as a gambling site. The offending URLs were "slot booking system for property registration", coverage of a politician named Pawan Kalyan, and articles about Escorts Kubota, which makes tractors. A fashion trade title was flagged for covering Kalyan Jewellers, a real jewellery brand. A Canadian local newspaper was flagged for its own crime reporting: police seizing oxycodone, a child pornography prosecution, a charity poker run, and a council debate about casino revenue.
Every one of those is a real publication doing its job. A keyword list would have rejected all three.
Two rules fix most of it. Archive-dated URLs, the /2013/05/19/ pattern, are newspaper archives rather than paid placements. And what matters is not that a page mentions gambling but whether it links to an operator. We fetched the flagged pages and looked: 206 of the 3,805 domains link directly out to betting sites. Those are the unambiguous ones.
When we asked a model to read the flagged pages and decide whether they were editorial coverage or monetised placements, it called 625 editorial and 723 monetised. Without that step we would have thrown away 625 real publications.
Check three: does the site have a subject
The failure that catches people out has no gambling in it at all.
Take a site with a genuine tool or a genuine local audience, whose blog publishes, in order: a piece about a therapy device, one about pest control, one about medical billing compliance, one about mini-split condensate drainage, one about end-of-life planning. No single article is objectionable. Together they are not a publication, they are a queue of commissions.
Two numbers separate that from a real site. Posts per day, taken from the feed's own dates. And whether the recent titles share a subject.
Neither works alone. A local newspaper we checked publishes 22 times a day and a men's magazine publishes 76, and both are entirely real. High volume plus no subject is a mill. High volume with a subject is a newsroom. Low volume with no subject is a dormant blog taking commissions. On our shelf, that pair rejected 25 sites outright and sent 141 more for a human look.
What this does to the numbers
We have written before that we mirrored two backlink marketplaces, 159,736 listings, and would sell links from about 2 percent. That figure came from screening homepages.
Screening what the sites actually publish takes it to roughly 1 percent. 1,686 sites with a price attached, median domain rating 51, median price $455, median traffic 2,200 a month. We tightened the screen this month after a review of our own bought placements turned up sites the earlier version should not have passed, and we would rather publish the smaller, truer number than the flattering one.
The checklist, if you are buying this week
Before you pay for a placement, in about five minutes per site:
Open /sitemap.xml and search it for the flag words above. Look at the last twenty posts and ask whether they read as one publication. Open any flagged page and see whether it links out to an operator. Check the traffic against the authority, because a domain rating in the 70s on 200 visitors a month is a rebuilt corpse rather than a bargain. And check that the site's own name matches its domain, because trading on someone else's brand is common and invisible from a metrics table.
Then ask the question those checks are standing in for: would anyone visit this site if it never sold a single link?
What this proves, and what it does not
It shows that one in six sites which passed a homepage-level screen fail when you read what they publish, and that a naive keyword check would wrongly reject real publications at a similar rate.
It does not prove our current screen is right. It produced its own false positives on the first day, including flagging a mental health service as a gambling domain because "bet" appears inside its name. Every number here comes from one snapshot of two marketplaces plus one supplier list, in September 2026, and sites change. We re-screen continuously and publish what changes.