My backlink ledger says 20. Twenty links that are live, that I verified myself, on pages that are indexable, with anchors that carry no nofollow. That number felt like a real thing.
Then I asked a second question — how many different hosts are those 20 links on? — and the answer was 12.
Twenty links, twelve sources. Nine of the twenty sit on dev.to alone. One host, 45% of the "link count".
That gap is not a rounding error, and it is why I no longer report link counts to myself without the domain count next to them.
The split, honestly
Here is the whole thing as of today, one row per host:
| host | followed links |
|---|---|
dev.to |
9 |
twelve.tools |
1 |
tryrocket.ai |
1 |
launchvault.dev |
1 |
junkstartups.com |
1 |
indielaunch.dev |
1 |
dledex.com |
1 |
aitoolseekers.com |
1 |
alterbase.co |
1 |
aigamehub.app |
1 |
coperogames.mataroa.blog |
1 |
coperogames.wordpress.com |
1 |
Twenty links, twelve hosts. The top row is 45% of the total and it is one domain.
And note the second thing hiding in that table: three of the twelve are my own blogs on free platforms (mataroa.blog, wordpress.com, plus dev.to which is a community account). Those are real pages with real followed links, but they are not independent editors deciding my game is worth linking to. They are me, publishing.
Why the count lies
1. Link equity is per host, not per page. Search engines do a lot of work to avoid counting many links from one site as many recommendations. Whether the adjustment happens at the host level, the template level, or the paragraph level, the effect I can observe is the same: the tenth article on the same blog does not feel like the tenth source.
2. One policy change deletes the whole row. Nine links on one host is nine links that die together if that host changes its outbound rules, suspends the account, or starts adding nofollow to post bodies. Nine links with nine different failure modes would be a different situation. Nine links with one failure mode is a single point of failure wearing a big number.
3. The count hides the shape of the work. Nine of those twenty came from writing. Two came from roughly twenty-one directory submissions. The rest were older finds. A single number — "20 backlinks" — makes those three completely different activities look identical, and they are not: one is repeatable in an afternoon, one took three days of forms, and one I cannot reproduce because I do not remember what I did.
4. "Live" is a snapshot, not a fact. Every one of these 20 is a measurement with a date on it. A link that is followed today can be nofollow next Tuesday when the site ships a redesign. Counting links without an expiry date is counting promises.
The two conversion rates, which is where it gets uncomfortable
Put the two channels side by side, same project, same weeks:
| channel | attempts | live + followed | rate |
|---|---|---|---|
| directory / launch-site submissions | ~21 | 2 (dledex.com, aitoolseekers.com) |
~10% |
| articles on platforms we control | 9 | 9 | 100% |
The 100% is not a victory. An article on a platform I control produces a lived link essentially every time, because I am the editor — the only gate is my own publishing form. What the 100% actually measures is my ability to publish, not the value of the link.
The ~10% on directories is the more informative number, because every one of those ten failures has a reason: a paid tier, a nofollow template, an account requirement, a review queue that never moves, a page that never gets created. Reasons can be acted on. A 100% rate has nowhere to improve.
So: the content line wins on determinism, the directory line wins on independence — a directory that links to me is a third party saying the game exists. Both are in the ledger, and I would rather have the honest split than a blended number that hides which is which.
The three numbers I track now
Per project, three numbers, no blended score:
- Submission success rate — of the things I tried to submit, how many produced any public page at all. This is about the form, not the link.
- 7-day live + follow survival — of the links that went live, how many are still live and still followed seven days later. This catches the template changes and the "we deindex new listings after a week" behaviour that a day-one check never sees.
- 30-day referral sessions — did anyone actually arrive. A followed link that sends zero humans for a month is a very different asset from one that sends thirty.
None of those three can be faked by publishing more articles on the same host. That is the point of choosing them.
What I'd tell someone one month in
- Record the host at the same time you record the link, and count distinct hosts as your headline number. It is one extra column and it changes the decisions you make.
- Do not chase 100% conversion on a channel you control and call it validation. Write down which numbers are "can I publish" and which are "will a stranger link to me".
- Keep the failures with reasons. My rejected list is more useful than my live list, because ten rejections with reasons is a model of the terrain, and twenty live links with no rejections is a list of things that happened to work once.
- Put a date on every verification. A backlink record without a date is a memory, not data.
I build Copero, a free browser football career game — no download, no account. The 20-link / 12-host split, and the ~21-submission / 2-link directory batch, come from the project's own ledger on 2026-11-20, verified page by page rather than taken from any dashboard.
Top comments (0)