DEV Community

Copero
Copero

Posted on

I checked 147 backlink records. Nine of them actually pass a link.

I keep a backlink ledger for my browser game. 147 rows, filled in over two weeks.
Each row says what I did: submitted, live, rejected.

Last night I re-checked every row that claimed to be live, by fetching the page
and reading the HTML. Not the dashboard the directory shows me. The page.

Eight of the twenty still pass a link.

That number is not the interesting part. The interesting part is what the other
twelve are doing, and what the 115 rows marked "submitted" turned out to be.

The ledger, as it stood

Ledger status Rows
live 20
submitted 115
rejected 12

I trusted that table because I wrote it myself. That turned out to be the problem.

Twenty rows said live. Eight are

A row is "live" because at some point a page existed. It says nothing about
whether that page passes a link, whether crawlers may index it, or whether the
canonical tag points at my listing. So I fetched all twenty and applied the three
checks again, one URL at a time:

  • an anchor pointing at my domain, with no nofollow, ugc or sponsored;
  • the page itself indexable (no noindex in meta robots or X-Robots-Tag);
  • a canonical tag pointing at the page I was reading.

Result:

What the twenty pages turned out to be Count
Anchor, indexable, self-canonical 8
Anchor, but nofollow / ugc 7
Anchor looks fine, page carries noindex 1
403 to every client I have 4

Eight out of twenty. Forty percent.

One detail worth its own paragraph: I first counted nine, and the ninth was a
Bear Blog page whose link had no restrictive rel at all. Clean anchor, HTTP
200, self-canonical — and a noindex in the head. Worthless, and invisible if
you only check the anchor. That is the difference between counting links and
checking pages.

Two targets on one host can disagree

Easy mistake I made: grouping by domain. dev.to appears twice in my ledger —
my profile page and a build article. The profile link carries rel="me ugc".
The article's link carries nothing restrictive.

Group by host and you get one row, one verdict, and you are wrong about the other
page by definition. Judgement has to be per page, because a link is a property of
a page, not of a domain.

The 115 rows marked "submitted"

This is where it gets uncomfortable.

Thirteen of them had a URL. I fetched those thirteen: ten are live now. None of
the ten passes a link — five carry nofollow, four have no anchor to my domain
at all (the page exists, my game's name is on it, the link is not), one is
noindex.

The other 102 rows have no URL in them at all.

That means those 102 submissions cannot be verified. Not today, not next week.
There is no page to fetch, no anchor to read, no way to answer "did this work".
All I know is that I filled in a form and something happened afterwards.

I built a prober that tries the plausible listing paths on each of those 102
sites — /<slug>, /games/<slug>, /products/<slug>, site search — and looks
for a link back to my domain. Some will resolve. Many will return "not found"
against a site that never published anything.

What the rows have in common

Reading back through the notes, the failures cluster:

  1. No URL recorded (102 rows). A submission without a public URL is a submission you have decided not to verify.
  2. A requirement discovered too late (9 rows). "Free listing" that needs a reciprocal badge in my footer, or a paid tier to remove the nofollow.
  3. The page exists but links elsewhere (4 rows). The directory hosts the game inside its own page, so there is no outbound anchor to earn.
  4. A login or human check I could not pass (5 rows). Three of those are interactive Cloudflare challenges on the submit form.

None of these is a directory being dishonest. Each is a step where I recorded an
intention and told myself it was an outcome.

The rules I use now

Verify the page, not the row. One anonymous fetch, then read rel, the
robots meta and the canonical. If it 403s, write "403" and move on — do not round
it up to live.

Record the URL at submission time or write "no public page". That one field
is the difference between a ledger and a diary.

Judge per page. Never per domain, never per site owner.

Count follow, not live. I had 147 entries and 9 dofollow links. Fifty rows of
"live" that pass nothing look like progress on a chart and are not progress.

The ledger is still worth keeping. But the useful version of it is smaller,
slower to fill, and full of URLs I can click.


I build Copero, a free browser football career game — no download, no account:
https://copero.games/?utm_source=devto&utm_medium=blog&utm_campaign=backlink-2026-11
The re-check above is a script: fetch, parse anchors, read robots and canonical,
store the answer next to the URL. If you keep a submission list, that is the
half-hour that makes it mean something.

Top comments (1)

Collapse
 
ref_vid profile image
Louis Chi •

"Count follow, not live" is the rule we ended up with too. One more column changed our count again: whether Google has actually indexed the listing page. A clean followed anchor on a page that never gets crawled still counts for nothing, and directory listing pages are thin enough that plenty of them never make it into the index. A site: search for the exact listing URL is crude but free. For the four 403 rows, it's worth opening each once in an ordinary browser before writing it off; some sites only block non-browser clients, so the link can be real even when your script can't see it. Of the nine, how many turn up in a site: search today?