Deleted in Five Days: Our Wikidata Item Didn't Survive
On October 12, 2026, we created a Wikidata item for a home-appliance factory we work with — part of the entity-building step in our GEO process. On October 17, five days later, an administrator deleted it: the item did not meet Wikidata's notability policy. The item was not vandalized or disputed. It was simply, by the policy's own criteria, not notable enough to exist. This post is the post-mortem: what the policy requires, where our item stood against each requirement, and the rules we now work by.
What we did
The reasoning seemed sound at the time. The factory is real: an established manufacturer in Cixi, Zhejiang, selling household appliances to overseas buyers under its own brand, with certifications and years of production behind it. It has an official website stating its facts. Wikidata is one of the sources AI systems draw entity knowledge from, so giving the factory a structured item — name, industry, location, official site — looked like basic entity hygiene, the knowledge-graph equivalent of a correct directory listing.
We created the item with statements sourced to the factory's own website, and considered the job done. Five days later the item was proposed for deletion and deleted. The stated reason was short: it did not satisfy the notability policy. No argument about accuracy — nobody claimed the factory does not exist. The claim was that existence is not the bar.
What the policy actually requires
Wikidata's notability policy gives an item three ways to qualify, and an item needs only one of them. First, a valid sitelink: the item is attached to a page on a Wikimedia sister project — a Wikipedia article, a Wikisource author page, and so on. Second, it describes a clearly identifiable entity that serious, publicly available references describe — and the policy is explicit that self-published material does not count as a serious reference, however large the outlet: a company's own website, its press releases, and its advertising are not independent evidence about itself. Third, structural need: the item is required so that other, already-notable items make sense — the classic case is the founder of a notable company. A cluster of non-notable items linking to each other does not create notability.
Administrator practice adds a working threshold on top of the text: roughly, two independent media pieces in which the entity is the subject — featured, not merely quoted or mentioned in passing — make an item effectively undeletable. Coverage has to exist before the item does, and the item is expected to carry its evidence from the day it is created: references attached to statements, and independent coverage linked directly. A bare item that plans to earn its notability later is reviewed within days, and reviewed harshly.
Where our item stood
Against criterion one: no sitelink. The factory has no Wikipedia article, no Wikisource presence, no sister-project page of any kind. Against criterion two: the entity is clearly identifiable, but the only references we attached were its own website — precisely the class of source the policy excludes. There was no independent media coverage in which the factory is the subject; not two pieces, not one. Against criterion three: no structural need — no notable item depended on ours to make sense. Zero for three. The deletion was not a moderation accident. Read against the policy, it was the correct outcome, and it was predictable before we pressed create.
There is one more rule worth stating, because it disciplines what happens next: do not simply recreate a deleted item. A recreated item is treated as an evasion, deleted faster, and can put the account at risk. The legitimate route is an undeletion request to the deleting administrator, with concrete evidence links attached — evidence we do not currently have.
The rules we now work by
Entity building runs in the opposite order from the one we used. First comes the off-Wiki record: independent coverage in which the entity is the subject, earned the slow way — trade press, real interviews, third-party profiles. A second route exists for authors: publish original work openly, have it collected by Wikisource, and the author page that follows is a valid sitelink under criterion one. Only when one of the three criteria is genuinely met does the item get created — and it is created fully dressed, references on its statements and independent coverage linked from day one.
The deeper lesson is the one this whole site keeps relearning. There are no self-declared shortcuts in AI visibility. A vendor cannot cite itself into being recommended, a site cannot manifest itself into being cited, and a company cannot register itself into being notable. In every case the verdict is assembled from independent sources, by systems — human or machine — that discount exactly the kind of evidence the subject controls. The item took an afternoon to create. The notability it needed takes quarters to earn. That order is not bureaucracy; it is the mechanism working as designed.
Sources & method. Dates and the deletion reason are from our own work log: item created 2026-10-12, deleted 2026-10-17 under Wikidata's notability policy. The three criteria and the exclusion of self-published sources are from the policy text of Wikidata:Notability; the two-independent-features threshold and the undeletion route reflect published administrator guidance on the same project. The factory is anonymized as a home-appliance manufacturer in Cixi, Zhejiang; no company name is given.
© 2026 Jun Liu · JUN & GEO. This article is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License (CC BY-SA 4.0) — https://creativecommons.org/licenses/by-sa/4.0/. You may share and adapt it with attribution and under the same license.