A seller picks the rating.
They do not pick the rest.
A domain rating counts links, and links can be acquired. What a seller cannot choose is whether Google sends anyone, which queries the site ranks for, who links to it, or, from the public web graph, how many outside domains the site links to compared with how many link in.
We measure all five. A domain has to pass every one of them before it reaches a plan.
the five tests
Checking traffic catches one in eight.
Traffic is the test everyone knows and the one a buyer can run for themselves. It accounts for a small fraction of what we reject. The rest have visitors and fail somewhere less visible, including two tests no rating tool runs: whether a subdomain has any footprint of its own, and whether a publisher links out to five times as many domains as link in.
No readers
Real link authority, and effectively nobody arriving from search. The classic hollow domain, and the one test you could run yourself.
No footprint
Authority and some traffic, yet it ranks for almost nothing. Whatever brought those visitors, it was not Google trusting the site.
Bad company
The rating is real, but it was built out of links that are themselves manipulated. Authority inherited from a network, not earned.
Bad ratio
Links out to 500 or more outside domains, five times as many as link in. A placement business with a publication wrapped round it.
Borrowed rating
A subdomain sold at its parent's rating, with no search visibility or link authority of its own. The number belongs to someone else.
A domain can fail more than one, so these do not sum to 49,669.
the finding
The forgeries are concentrated where the budgets are.
You would expect the fake rate to climb steadily with claimed authority, because a bigger lie is worth more. It does not. It peaks at DR 90–100, where 50.2% of domains fail, then falls away above DR 70.
That band is not chosen at random. It is the workhorse range most link budgets are spent in: high enough to be worth buying, low enough not to invite a close look. Forgery follows demand.
share of domains flagged · domains measured at right
where they come from
Nobody is checking the inventory.
We do not name suppliers, but we can tell you how they operate. A self-serve marketplace lets anyone list a site. A managed seller has a human decide what gets sold. That single difference predicts the fake rate better than price does.
Both kinds sell real inventory too, which is exactly the problem. You cannot dodge the fakes by dropping a supplier. You can only measure the domain.
what we do with it
Flagged everywhere, excluded by default.
On every response
Each domain carries a verification state, its measured organic traffic and its spam score, next to the figure the seller claimed. You can see where they disagree instead of choosing one on faith.
Out of plans automatically
Ask for a campaign and suspect domains are left out unless you explicitly ask for them. The failure mode of the other default is that you find out in three months.
Never as a verdict
It is a flag, not a judgment, and we publish it as a share of the catalogue rather than against a named domain. A legitimate site can be missed by an index, and a publisher did not consent to being called fake on our website.
questions
How the measurement works.
What makes a domain suspect?
Any of five tests. Substantial link authority with effectively no organic search presence, or almost no ranking keywords: authority is built from inbound links, which can be acquired, while real people arriving from real queries cannot be. A spam score over 50 on the links themselves. A subdomain sold at its parent's rating with no footprint of its own. And, from the public web graph, a publisher that links out to 500 or more outside domains and five times as many as link in: a site that sells more placements than it earns links. Every response names which test fired.
Could a legitimate site be flagged?
Yes, and we say so on every response rather than presenting the flag as a verdict. A young site, a site that ranks mainly in a language or market the index covers poorly, or one that gets its traffic from somewhere other than search can all trip it. It is a strong signal that something needs a human look, not proof of fraud, and we never publish it against a named domain.
Why not just use the seller's DR?
Because the seller chose it. Most quote a figure from a third-party index, but which index, when it was measured, and whether the domain has changed since are all theirs to decide. We keep the claimed figure and record our own measurement alongside it, so you can see where the two disagree instead of picking one on faith.
Do you replace claimed DR with your own number?
No, deliberately. Our independent measure runs on a different scale and correlates with claimed DR at only 0.36. Substituting it made every measured site clear a DR60 filter and silently tripled the apparent DR50+ inventory. Verification's job is the quality flag, not quietly redefining the scale you shop on.
Can I buy suspect inventory anyway?
Yes. It is excluded from plans by default and returned in searches with the flag attached. An expert may have a reason we do not know about, so it is a default rather than a wall.
The full method, and the same measurement broken out by subject, supplier model and placement type, is on the market report. It is free and it does not need an account.
Check a domain before you buy it.
Every measurement above is a field in the API, on a free key, with no card. The result for a specific domain is not published here: call GET /v1/sites/{domain}, or read the API reference.