We check a site three separate times: before it joins, the moment a link is submitted, and every month for as long as that link is supposed to stay live. We fetch the pages ourselves. Nobody's word is taken for it, not even ours.
Every site is crawled and checked before it joins, and re-checked while it stays. The bar is not about size (small local sites are welcome and are much of the point). It is about whether the site is real.
Authority that does not match traffic
For a site whose traffic should track its authority, a high domain rating with almost no organic traffic means links were bought rather than earned. DR 40 or above with under 500 monthly organic visits is a flag, unless the site is a reference publication, which is a separate case below.
A traffic cliff
Organic traffic down 50% or more over six months usually means an algorithmic or manual action. Sites in that state are declined until they recover.
Reference sites are judged differently
Scientific and medical literature, standards bodies, technical references, legal commentary: sites that are widely cited within a narrow field and have modest traffic because of it. These are exempt from the traffic rules above, on a moderator's decision rather than automatically. Being authoritative and niche is not suspicious, and treating it that way would exclude some of the best citation targets there are.
Set during review; never inferred from the numbers alone.
A real business behind it
A site with no named authors and no real business behind it does not get in. We check for consistent about, contact, and address information matching the business the site claims to be.
Coherent subject matter
A "blog" covering travel, finance, CBD, and software is not a publication, it is a place to put links. Past 4 unrelated categories, a site goes to a person and most do not get in. Writing about a neighbouring subject is not sprawl: nobody is limited to their own category.
A real publisher is the exception and a reviewer grants it by hand. A trade journal or a local paper is allowed 8, because breadth is the job. Never inferred from the breadth itself.
Clean outbound linking
We look at who a site already links to and how. A travel article carrying an anchor like "SEO consultant services" is the paid-link fingerprint, and it is disqualifying on its own.
Not a known bad neighbour
Templates, analytics IDs, hosting neighbourhoods, and registrant patterns from previously rejected sites are remembered. A new applicant matching one is flagged automatically.
A declined site can reapply
A decline is not held against you forever. Release the domain from your account and add it again after the same 30-day wait every re-added domain goes through, and it is reviewed again on its own merits. The decline note says what to fix in the meantime.
Nothing is taken on trust. Every claimed placement is fetched independently, whatever tooling reported it, including our own WordPress plugin, which can insert a link but can never confirm one.
The link resolves to exactly the assigned URL
No shorteners, redirects, or tracking wrappers in between.
It is followable and on an indexable page
No nofollow, sponsored, or ugc. The page returns 200, is not blocked by meta robots or X-Robots, and the link appears in the served HTML rather than only after JavaScript runs.
It sits in the article body
Located by content extraction, not by position on the page. Header, footer, sidebar, author bio, comment, and blogroll all fail.
The anchor matches what was assigned
The assigned anchor or an approved close variant. Anything else fails, and a swap after the fact is caught at the next monthly check.
The paragraph genuinely supports the citation
The paragraph containing the link is compared to the passage it cites, and read against a rubric: does the sentence make a claim the target actually supports? Would it survive without the link? Does the topic transition make sense within the host article?
It is not boilerplate
Citation sentences are compared against recent placements across the whole network. Near-duplicate phrasing is rejected. Identical patterns across sites are exactly what makes a network visible.
A link that disappears three weeks after it earned tokens was never worth anything. We keep checking placed links, and the tokens follow the link.
Re-checked monthly for 6 months
Every verified link, every month, against the same standard it passed the first time. A link switched to nofollow, moved to the footer, or given new anchor text counts as dead.
14 days to fix it
If a monthly check finds a link dead you are told, with what changed, and have 14 days to fix it. Restore it inside that window and nothing is taken back; miss it and the tokens go.
90% of your links must stay live
Below that, receiving pauses until the rate recovers. You can keep giving and keep earning in the meantime. The freeze is on what you can receive, not on what you can give.
Taken-back tokens can put you below zero
If the tokens for a dead link were already spent, the balance goes negative and receiving is blocked on every domain on the account until new links earn it back above zero. What is owed does not clear on its own.
A removed domain's links are still watched for 90 days
If you stop working on a site, you can remove it: its place on your plan frees up, it leaves the network, and its history stays on your account. Links it placed in the 90 days before you removed it are still checked, and their tokens can still be taken back. Otherwise the obvious move would be to earn on a domain and let it go the week before the links come down.
A suspended site's inbound links are honoured
If a site is removed from the network, everyone who linked to it keeps their tokens. We punish the bad site, not the people who linked to it.
Next: Connect from WordPress or an AI assistant →