DEV Community

Copero
Copero

Posted on

I submitted 21 backlinks. Five were real.

Twenty-one backlinks. That is what my submission ledger said was live for a small
browser game I build in my spare time. Directories, launch sites, startup lists —
the usual round of "submit your product and get a link".

Then I fetched every one of those pages anonymously, with no cookies and no
login, and read the HTML instead of the dashboard.

Five survived.

This is what I found, how I checked it, and the three rules I now follow before
submitting anywhere.

What "live" means to a directory

A directory marks your submission live when the page renders. It does not check
whether the link carries rel="nofollow", whether the page can be indexed at
all, or whether the page's canonical tag points at your listing or at something
else entirely.

That gap is where most of the value of a directory link goes to die. The listing
is real, the page is real, the anchor text is real — and the link still passes
nothing, because the page it sits on will never be in an index, or because the
page is a duplicate of a URL that will.

The numbers below are from 21 submissions for one product. Not a study of the
whole directory industry — just what actually happened to me.

The three checks

I stopped trusting statuses and started checking three things on the page itself.

1. rel is a whitelist, not a blacklist. A link passes only when the anchor
carries no rel that restricts it. nofollow, ugc and sponsored all fail;
an anchor with no rel attribute at all passes. Redirects in between do not
matter for this check, but an anchor that is a JavaScript handler or an image
does — there is nothing for a crawler to read.

2. Can the page itself be indexed? A noindex in the meta robots tag or in
the X-Robots-Tag header makes the whole page worthless, no matter how good the
link is. This one is easy to miss because the page looks completely normal.

3. Does the canonical point at itself? If <link rel="canonical"> points to
a different URL, the page is telling search engines "this is a duplicate of that
one". The link belongs to whichever URL is canonical, and if that is not your
listing, your anchor is decoration.

I also recorded HTTP status, whether an anchor existed at all, and whether a
login wall appeared. A 403 got one retry with curl — some sites reject one
client and serve another, and I did not want to call a page dead because of my
own user agent.

The results

Twenty-one submissions. Sixteen left a public listing page. The other five were
submission forms rather than listings — nothing to fetch, so they are not counted
in anything below.

Of those sixteen:

What the sixteen pages showed Count
Anchor, dofollow, indexable, self-canonical 5
Anchor, but nofollow / ugc 3
Anchor present, HTTP 200, but the page cannot pass value (noindex, or a canonical pointing elsewhere) 2
Page exists, no anchor at all 3
403 to every client I tried 2
Unreachable 1

(5 + 3 + 2 + 3 + 2 + 1 = 16.)

Five out of sixteen. Thirty-one percent. And the five that passed were not the
ones I would have guessed: two were small community-style sites, one was a blog
platform, one a "made with" showcase, one a niche game directory.

The part that surprised me

Three pages looked perfect and passed nothing.

One had a dofollow anchor with two occurrences of my product name, and a
noindex in the head. The most generous-looking link of the batch was worth
zero. Another had four anchors — every one of them carrying nofollow ugc — and a
canonical pointing at a URL I do not control. Two failures in one page: the
links do not pass value on their own, and the page is, in search-engine terms,
a copy of something that is not mine. A
third returned 403 to every request, including a different client — the page may
well exist for humans, but if I cannot see it I cannot claim it.

The lesson is uncomfortable: a link that looks better is not a better link.
Anchor count, placement, and a friendly "do-follow!" badge on the directory's
pricing page tell you nothing about the two attributes that decide whether the
link counts.

The counters disagree with each other

There is a second ledger in this story: 20 links my own running list called live,
for the same product, counted with looser rules. It overlaps the batch above —
twelve of the sixteen pages are the same sites — so this is not extra coverage, just a
different way of counting. On those 20:

  • the site owner had reported 14 followed links;
  • counting the way a directory does — anchor present, page returns 200, and the site's own label says follow — I got 10;
  • requiring all three checks, I got 8.

The distance between them has a shape. The two rows between 10 and 8 are a
profile page whose anchor carries ugc and a post whose page carries a noindex
in the head. The four between 14 and 10 are the four pages I have never been able
to open — they return 403 to every client I have. My oldest ledger counted those
four as live because the submission had been accepted, not because I had seen the
page. That is an inference from my own bookkeeping, not a measurement: 14 = 10 +
the four I could not check, and the four are exactly where the optimism lived.

That is the part that made me stop using dashboards for this. Every one of those
counts was produced honestly. They just measured different things.

The same lesson one level up

After the verification pass I went the other way: instead of checking links I
already had, I built a small checker for the directories themselves, to find
places worth submitting to. Four questions per site: is the submission form
reachable, does it need a human check (CAPTCHA), is there a free tier, and does
the site actually take a product like mine.

Seven sites passed all four checks. Free, open form, no CAPTCHA, category that
matched.

I opened all seven in a browser. Two wanted money before the listing existed —
one at $79, one at $19.99. Three put a CAPTCHA in front of a free submission
(two image codes, one reCAPTCHA). One accepted blogs only. One required an
account, and its sign-in page returned 403 to an automated browser.

Zero submitted. The static checks were not wrong about what they measured; they
were measuring the wrong layer. "Free" on a directory almost never sits on the
submit button. It sits behind a login, behind an upload endpoint, or in a
sentence two paragraphs down that says a link back is mandatory.

Three things I do differently now

1. Verify the page, not the dashboard. One anonymous request, then read
rel, noindex and canonical. It takes a minute per link, and it is the only
number I now trust. If a page 403s, say so and move on — do not round it up to
"live".

2. Treat "free" as a claim to test. Look for a price next to the word
listing. Look for the account requirement before filling the form. Grep the page
for "link back" — a mandatory reciprocal link is not a free listing, it is a
trade you did not agree to.

3. Keep the URL at submission time. If a submission does not leave a public
URL behind, you can never verify it: not next week, not next month, not when
somebody asks whether it worked. "No public page" written down honestly is worth
more than a hopeful status you cannot check.

What I would tell another indie developer

Directories are not worthless, but they are not a channel either. In my batch the
five real links came from places that had an actual audience for the product — a
game directory, a showcase, a community — and the worthless ones came from
directories whose business is selling listings.

The links that keep working are the ones somebody chose to write. That is slower.
It also does not need a dashboard to tell you it happened.


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
I write about building and measuring it. The link checks above were done by hand
on 21 pages; if you want the exact method, the three tests are the whole trick.

Top comments (1)

Collapse
 
posting-dude profile image
Posting Dude •

The three pages that looked perfect and passed nothing is the part that stuck with me, especially a dofollow anchor sitting on a noindex page. Did the five that passed share anything you could spot before submitting, or is fetching the page afterwards the only reliable check?