Zero impressions with eight of eight pages indexed
· landing · discoverability
The question was simple: why does pipfold.com get zero impressions and zero
clicks in Google Search Console?
The answer was not, but it is measurable in five minutes — if you know which two numbers to put side by side.
First: is it technology at all?
pnpm nachfrage pipfold.com 90
→ 0 impressions · 0 clicks · 0 days with data in 88
sichtbarkeit-messen.mjs --stichprobe=8
→ 8 × "submitted and indexed"
site:pipfold.com
→ all five languages, manual, devlogThat disposes of the entire obvious suspicion. Robots, sitemap, hreflang, canonical, load time — any of those would explain not indexed. For indexed and still zero they explain nothing.
Indexed and zero means: the page ranks for nothing anybody searches for. Whoever starts checking technology at this point loses a day.
Cause one resolves by itself
The domain was bought on 14 August. On the day of measurement it was nine days old. Zero in a ninety-day window is not a diagnosis at nine days of existence; it is arithmetic.
That eight of eight pages were already indexed is, for nine days, actually fast.
Cause two does not resolve by itself
The page did not name its own subject.
“Shut the Box” appeared twice in the entire HTML of the front page — once
in the <title>, once in og:title. In the visible text: never. The H1 read
“Nine flaps, two dice.”
That is a fine line. It is also a phrase nobody searches for. And the brand name
that stood at the front instead is invented — pipfold has zero search volume
by construction, which was in fact the selection criterion when
renaming.
The most expensive line was in robots.txt
# The play page itself is an empty canvas with a 14 MB attachment — it does
# not belong in the index, the landing page does.
Disallow: /spielen/That justification was right once. It applied to the 5.8 KB Godot shell of the time, which contained not one sentence of text.
Measured on 23 August:
curl -s https://pipfold.com/spielen/ | wc -c → 4306843 KB, with a <title>, a description, loading text and the word “mana”
twenty-seven times. The justification had expired; the block had not.
And it cost exactly the strongest search intent. “shut the box online spielen” wants a page that plays, not one that describes — and the only page that plays was not allowed in the index.
This is the most unpleasant kind of legacy: a comment that justifies the decision cleanly, and which is therefore never re-examined. Whoever reads it finds it convincing. It merely describes a state that no longer exists.
Demand is split up; one page does not catch it
Before any recommendation came the obligation to look at the actual results page. For the three strongest queries, German and English:
| Query | AI overview? | Who is on top |
|---|---|---|
shut the box online spielen |
no | shutthebox.fun, playonlinedicegames.com, spiele-kostenlos-online.de |
shut the box regeln |
no | Wikipedia, then a family blog and three manufacturer PDFs |
play shut the box online free |
no | the same sort of thin browser-game sites |
No AI overview above the web results — so the question is not answered away above every listing. And the competition is thin.
More interesting than the listing was the “People also ask” box. It states the work list verbatim:
rules 2 players · rules 4 players · 10 rules · 12 rules · doubles rule · instructions PDF · Shut the Box in German
Seven questions, all answerable — and all of them folded until then into a single page called “Manual, questions, origin”. That heading names neither the game nor the player count nor the flap count. It catches not one of them.
What was built
Three separate pages, in German and English each: /rules with one round
played all the way through, /variants with 1–9 / 1–10 / 1–12,
/multiplayer with two to four players and the tournament.
Plus front pages with the searched term at the front and the brand name at the
back, and structured data for the first time ever — before this, no page on
this site carried an ld+json.
Rejected: leaving the three sections in the manual and building pages from them as well. That would have been the comfortable route: break nothing, only add. But two pages carrying the same five rules do not rank twice — they rank against each other. The sections moved out; they were not copied. The manual keeps what nobody searches for and somebody still needs: controls, limits, signature, origin.
And also rejected: twenty doorway pages. “shut the box rules 3 players”, “… 4 players”, one address each — formally the demand would support it. Three pages with their own content beat twenty that copy from each other, and that difference decides whether all three end up ranking or none of them do.
And a lock nobody had opened
Finally, a finding that has nothing to do with framing: the sitemap sat
correctly at /sitemap.xml, robots.txt pointed to it — but it had never been
submitted in Search Console. sitemaps.list answered {}.
Submitting was not possible: our token carries only webmasters.readonly, and
sitemaps.submit answers 403 ACCESS_TOKEN_SCOPE_INSUFFICIENT. That is a click
in the interface that a human has to make.
The lesson is not about SEO. It is: a comment that justifies a decision ages worse than the decision itself. The robots.txt convinced everyone who read it for four weeks — because it was well argued and because nobody re-measured the number in it.
Two seconds of curl … | wc -c would have been enough.