A post published on 17 August by Weird Gloop, the outfit that hosts runescape.wiki and minecraft.wiki, reached the front page of Hacker News in the early hours of this morning and passed 400 points by the afternoon. Its claim is narrow and its implications are not. Since Google’s March 2024 core update, according to people running some of the largest independent wikis on the web, a site launched on a brand-new domain gets exactly one page into Google’s results: the homepage. Everything else is crawled, stored, and never shown. If you have a product launch planned on a fresh domain, that is a line in your budget nobody has costed.
TL;DR
- Weird Gloop reports that roughly 90% of wikis launched on brand-new domains since the March 2024 core update have had only their homepage appear in Google results, sometimes for up to a year.
- It is not a duplicate content problem, not a ranking problem, and usually not an indexing problem: the pages are crawled and appear under a
site:query, they just never surface for real searches. - Wikis launched on subdomains of already-indexed domains are unaffected, and the parent domain does not need to be well known.
- Search Console will report those pages as indexed, so the failure is invisible to standard monitoring. The diagnostic is indexed page count versus the number of distinct pages receiving impressions.
- Google has never acknowledged a sandbox and the evidence here is observational, but the cheap mitigation, launching under a hostname Google already knows, costs nothing to adopt.
What the wiki operators documented
The post, written by the operator who goes by cookmeplox, describes a pattern across roughly ninety per cent of the wikis they know of that launched on a brand-new domain after the March 2024 core update, which began rolling out on 5 March 2024 and was confirmed complete on 26 April. Named examples include gta.wiki, hytalewiki.org, hollowknight.wiki and the official Path of Exile 2 wiki, plus a couple of dozen smaller sites. All were serving normally when we checked them on 8 September 2026. Around 85% of game wiki traffic arrives from Google, so a site that can only surface its front page is, in the author’s word, catastrophic.
Four details make this more interesting than the usual grumbling after a core update.
It is not duplicate content. The pattern holds for wikis covering brand-new subjects with no rival copy anywhere, which the author confirmed directly when challenged on it in the Hacker News thread. It is not ranking either: in several cases the homepage outranks the incumbent’s homepage, and it is still the only page from the domain that appears for anything. It is usually not even indexing, because the rest of the pages are crawled, stored and retrievable with a site: query. And it is not permanent. Hollow Knight’s wiki went in around March 2024 and came out roughly nine months later. Some sites go in and out. Some escape when a game update sends a burst of new traffic and content their way.
The most useful observation is the trigger. It does not track the registration date of the domain. It tracks roughly when Google first indexed it. You cannot park a domain for two years and buy your way out.
The failure mode no dashboard reports
Every tool you would normally use to check whether a launch is working will tell you the launch is working. Search Console’s page indexing report will show your pages as indexed, because they are. Your crawler will report clean status codes and valid canonicals. Your sitemap will be accepted. Your log files will show Googlebot arriving and working through the site. Every technical SEO check passes.
What none of those surfaces is the gap between indexed and shown. A page that is indexed but never returned for a query produces exactly the same signal as a page that has not been discovered yet, which is to say no signal at all. The first thing anyone actually notices is a revenue or signup line that is softer than the model said it would be, three or four months after launch, at which point the working assumption becomes “our content is not good enough” and the team spends a quarter rewriting pages that were never going to be shown regardless of what was on them.
We made a version of this argument about Google’s AI Mode earlier this month: absence from a surface looks identical to not having measured that surface. This is the same shape, and it has a cheap diagnostic. Take the indexed page count from Search Console’s Pages report. Then take the Performance report over the last 28 days, group by page, and count the distinct pages that received at least one impression. If the first number is in the hundreds and the second is one, you have your answer. No standard dashboard computes that ratio, which is precisely why it is worth computing.
Domain history is an asset you cannot buy
Weird Gloop’s workaround is the interesting half. Wikis it launched on subdomains of established domains, including wiki.leagueoflegends.com and wiki.warframe.com, were all indexed normally and immediately. So was overwatch.weirdgloop.org, which was outperforming domains stuck in the jail seven days after launch. The parent domain does not need to be popular. It needs to be known.
That reprices three decisions teams usually make in about ten minutes. A greenfield product launched on a fresh brand domain now carries an unpriced discovery delay. A rebrand that moves a working site to a new name is no longer a redirect exercise with a short dip. Campaign microsites and vanity domains, the .ai someone buys the week before launch, are the worst case of all: short-lived sites that will spend their entire life invisible.
It also settles an argument the industry has been having for fifteen years. Subdomain versus subfolder used to be a debate about how link equity flows between hostnames. That is now the second question. The first is whether the hostname sits under a domain Google has already indexed, and if it does not, the equity discussion is academic because nothing but the front page will be returned anyway.
The honest caveats
Google has never acknowledged anything of the kind. “There is no sandbox” has been the official line since at least 2019, restated in 2021 as a description of algorithms working with very little data rather than a filter applied to new sites. Nothing here is confirmed behaviour.
The evidence is also observational and drawn from an unusually homogeneous population. Game wikis launch with thousands of pages at once, they are topically dense, and they typically arrive alongside an established copy of themselves. At least one operator in the thread reported a niche wiki launched about eighteen months ago ranking well within months. A twelve-page brochure site may never encounter this at all.
None of that changes the decision. If launching under a hostname Google already knows costs nothing, the confidence level in the diagnosis stops mattering very much.
What to do before your next launch
- Move the domain decision into discovery. It currently gets made late, by whoever is buying the name, on aesthetic grounds. It belongs alongside hosting and architecture, for the reason above.
- Launch on a hostname with indexing history where you have one. A path or subdomain under the existing company domain, with the vanity domain redirecting in, not the reverse.
- Plan the move as a migration, not a launch. Weird Gloop’s stated plan is to establish the wikis on subdomains and then 301 to the proper domains using Google’s change of address tool once they are settled. Treat that as a real project with a tested redirect map, the same discipline we set out in our piece on URL architecture.
- Instrument indexed-versus-shown from week one. The ratio described above, checked monthly, turns a silent failure into a number somebody owns.
- Budget acquisition for the gap. If a launch plan assumes organic traffic in month three on a domain Google first saw in month one, that is an unfunded assumption. Either fund paid acquisition for the window or change the launch domain.
The uncomfortable conclusion is that discoverability has become a property of the address you launch at rather than the site you build there, and that address is the cheapest part of the whole project to get right. Nobody is coming to tell you this has happened to you. The measurement is yours to build.
REPTILEHAUS works with founders and product teams on web architecture, launch planning and site migrations, and we would rather have the domain conversation with you at kickoff than in month four. Get in touch.
